00:01 Google Gemini is a breakthrough in AI with multimodal capabilities 02:14 Google's GEMINI is a family of models with different capabilities. 06:54 Creating a game called Guess the Country with clues provided 09:35 Gemini Ultra has surpassed GPT-4 in nearly everything. 13:52 Gemini's multimodal reasoning capabilities to understand and reason about users' intent 16:00 Gemini's capabilities in creating visually rich and interactive experiences 20:11 Gemini uses Advanced reasoning and multimodal capabilities to filter, extract and update large sets of scientific data 22:05 Google Gemini enables multimodal interactions with AI. 25:20 Google's GEMINI has advanced capabilities for identifying plants and generating text and image content. 27:00 Google Gemini exhibits advanced multimodal capabilities 30:15 Google's GEMINI will bring groundbreaking innovations in AI next year.
Google has had advanced AI for years. Anyone remember Google Duplex making AI assisted phone calls, presented at Google io in 2018? I think they would have integrated it into Google Assistant if people didn't freak out so much. Not that it was wrong to freak out. But now, Chat GPT has proven that many millions of people will incorporate AI into their lives. With all the competition, the Genie is out of the bottle. So, Google is like "alright! integrate everything we've got and hit them with Super Saiyan AI!" It's gonna be a bumpy ride folks!!
that is a problem because the current world government diaspora will only value you as a walking target for their new weapon advancement or a subject of radicalization so they have a new reason to go back to start of my reasoning. :P] The issue is not AI but that humans have no clue what to do with their free time that is constructive and not distructive. Food for thought. If you don't need to work and relay on fossil fuels will you still defend their use? //rhetorical question....maybe :P
To me, the significance of Gemini is that it all seems to be in one package. If it can indeed shift seamlessly between modalities, then this is a huge breakthrough and we aren't even in 2024 yet!
AI is definitely impressive! I have a few points though - 1. Will it not cause decrease in human interaction & dependency? 2. Relying on simple answers like is the omelette cooked, will it not decrease the basic common sense that humans are gifted with? 3. What are the controls that we have in place?
@@Indpendent01 Google could have used this in video instead! This is what I exactly mean, AI is great if used in ways that have a actual positive effect and not replace what we already do.
Yes, that's an excellent point. The time between iterations seems to be getting shorter and shorter; if you grew up with SciFi movies from the 80s (like Terminator) and you were living through the 90s / 2000s, you always thought "how is this going to happen by 2030 or something like that, the technology still seems so very far away". And now, just a few years later, it looks like all of what we saw before is "right around the corner". I can't help but feel excited about our future. 🙂
I hear you. Although I feel its the other way around. ChatGPT and others are forcing GOOGLE to release stuff they were gonna wait to release much later.
Yes because deadline pressure is exactly what you want when you're developing a technology that many experts believe could see the end of the human race.
@@AustinMulkaMusic The field of AI safety believes that the race condition caused by competition is dangerous. I believe them. The models have emergent properties: when you increase the size of capabilities, they do things we didn't know or intend for them to be able to do. Isn't that at least a little concerning? It doesn't have to be an extinction event to be bad. The first major application of AI (the social media algorithms) has already nearly torn society apart. It focuses on outraging people since it's the most effective way to harvest their attention, and exposes them to more and more siloed material. People are forced towards extremes, and we have more conflict and division than ever. They weren't programmed to do this; it was an emergent phenomenon. Now facts have become almost meaningless, and everyone lives in their own alternative reality curated by...the AI algorithm. So...we have clear evidence in the only major application that it can do enormous amounts of damage. It would be very silly not to be worried, and rushing is definitely, definitely not the answer.
Reliability is the question. Google has a history of not doing live demonstrations that were not staged to display the tech abilities. It's going to be interesting to see.
@@lutaayahuzaifahidris3679 Yes it WAAAAS a game changer, but the NEW MORE COMPLETE GAME NOW is GEMINI !!! ...On neutral basis Google always has had the most robust group of AI Researchers in the world since at least 12 years. Many of those were teachers of current Microsoft researchers.
Very impressive! But they will need to include a voice conversation mode ASAP to compete with open AI. Of course, other companies can use Gemini via API to produce services that will allow you to interact with an avatar that sounds like a real human.
For once the title isn’t click bait, this is actually shocking, I first thought wow then further into the video I thought WOW ! It’s not just keyboard input its video plus voice interaction, amazing !. If they could get this into something like Amazons Astro home robot it would be life changing for old folk or those living on their own etc, someone to talk with all day long, loving it
Until it's in my hands and I can use it to exponentially improve my productivity for a mere $20 per month like GPT4, it's just a hypothetical concept. I could care less about the top speed or handling of a car that isn't even on the market. Mine's plenty fast and I drive it every day. Let's do this again once we can all get our hot little hands on Gemini.
Google has a history of getting people excited for these things just to be silently cancelled. I doubt that’s the case here (because.. duh, it’s about their money), still. You right.
They do have integrated gemini pro in bard and i talked with it. It surely gave better and more detailed responses than its previous version. But the biggest downside for me to drop it was its accuracy. I asked it to give a timeline of all the ai developments that happened recently. It got the events right but not the dates. It said that gpt 4 was released in november 2023 and microsoft bonsai project in 2022. As a student i need reliable answers so bard is yet not suitable for me
Love everything except for the potential abuse in censorship for things Google deems "harmful" when it comes to actual discord, or research, especially with things like current events, politics, and hot button issues. Just don't be evil, and stay neutral, unless something is in violation of actual laws.
Their ethical code was "don't be evil" a while back. They've since removed that and not long after got into the dirty fossil fuel business. The latest which discuss me is the "project Nimbus" that they are involved in, read about it. But don't trust Google, for sure.
amongst the big companies, I trust Google the least to not be biased and try to big tech / big brother spy on people so I can't say I'm excited for any ai Google cooks up
Exactly. For instance, I noticed it did not incorporate ALL papers, but only the RELEVANT ones (19:48). Relevant to the subject, of course. But what if the system gets programmed to exclude certain studies that contain necessary information on a certain subject, but is deemed "not relevant" because it is not in accordance with a (biased) 'consensus'?
But we automatically assign value to everything we see or hear. We would not accept statements by Biden or Trump as equally true and equally important. No one can hold opposing views on every issue from energy to healthcare. There are 2 ways to cook chicken in an Instant Pot. One uses 1 cup of water and 10ish minutes high pressure. Another uses 4 cups water and 0 minutes. What would Gemini say if asked what's the best way to cook chicken in an Instant Pot? Combining the recipes would be a mistake. Picking the most popular would suppress novel approaches. What AI can't do is try out both recipes. Would it understand that one is a European method and the other is Asian?
This was my first prompt: Here’s a puzzle: two coins are on the table, touching. One is three times the diameter of the other. If I roll the smaller coin around the larger, maintaining contact, how many rotations will the smaller coin make when it has gone thru one orbit? Of course, Gemini was quite confident of its answer, but nonetheless, incorrect. It answered 5, complete with what looked like pretty convincing arguments and math.
You're explanation/instructions has room for improvements - it was correct in one of 50 different variations where u were not explicit enough in your request. You should check yourself into a soup kitchen line - Your job has now been replaced by Ai
To solve this puzzle, we can use the concept of the circumference of a circle and the relationship between the diameters of the two coins. Given information: - The larger coin has a diameter that is three times the diameter of the smaller coin. - The smaller coin is rolled around the larger coin, maintaining contact. Let's assume the diameter of the smaller coin is 'D' and the diameter of the larger coin is '3D'. When the smaller coin makes one complete orbit around the larger coin, it will make a number of rotations. The circumference of the larger coin is: Circumference of the larger coin = π × 3D The circumference of the smaller coin is: Circumference of the smaller coin = π × D To determine the number of rotations of the smaller coin, we need to divide the circumference of the larger coin by the circumference of the smaller coin: Number of rotations = Circumference of the larger coin / Circumference of the smaller coin Number of rotations = (π × 3D) / (π × D) Number of rotations = 3 Therefore, when the smaller coin has gone through one orbit around the larger coin, it will have made 3 rotations.
One thing I love about using GPT-4 via Bing, on my phone, is that it can verbalize the responses. And it does so with a very natural and human like voice in a accurate, concise and helpful way.
@@tranzhuman TOO low?...it's takes just seconds. why does it need to be faster?... how fast does it need to be?.... how fast is sufficient?... so many questions.
@@rw9207Copilot is great but you can't deny that it is slow compared to GPT-4. I put it to the test recently and the slow speed of copilot was unbearable compared to GPT-4
Unsurprisingly, like Chat-GPT, it hallucinates at times e.g. saying that plastics/rubber are less dense than water (in the duck part). The real danger of AI is that we'll struggle to tell what is real and what is fabrication/a distorted version of reality.
Something that will no doubt become common is running the same query in parallel across Gemini, GPT4, Llama, and coming up with a confidence score. We do this sometimes with human experts, like with radiocarbon dating where samples are sent to multiple labs and compared to establish confidence. In an ideal world everything would be checked by 3 humans, but that's impractical -- not so much with models.
@@jwoyaThis is exactly what everyone has to do with everything online especially anything that has been targeted to compete for our attention. I'm getting tired of doubting everything, because ulterior motives.
This whole talk about safty and filters reminds me a lot of 1984. Concentrate all the power in a single hand and than decided what is "safe" and what is not.
That's the whole thing right?, to use AI efficiently, you have to have some brain in the first place. AI enhances smart people but replaces dumb ones. Unfortunately it also prevents smart people from being nurtured in the first place. It creates and solves its own problems,leaving us with a net zero for work efficiency but a negative for society as a whole
A little bit of thinking, we already gave away with smartphones being essentially a part of us, a part of our brain and daily information gathering. Now we're at a point, where this outsourcing of complex thougts is really, really big.
I use it to search the internet and not have to deal with all the extra stuff, I also use it to rewrite things, like my resume. Some companies using AI to find new paths in medicine. It will be used to scan people to find disease. It will be used to make things work better. Video games with Ai monsters. Tech will boom.
So this multi-modal method to conception, observation, connection, and response combined with quantum computation which is making regular computers look like a mere calculator now...my mind is blown! I have never felt such exhilaration. Even if this kills me, I have lived to see water being pulled up from a well to this moment, so I would die having witnessed something very profound, unimaginable, and inspiring, but the beauty of it, or the horror of it is that humans created something far bigger than themselves. I used to question the reason for life, but now, I am silenced. This is some kind of proof of something I am too small to process or to which I can put words.
I noticed that GPT-4 has a higher score, specifically in commonsense reasoning for everyday tasks, with 95.3%, compared to 87.8% for Gemini. Does this mean that if we need AI as a tool that is useful and applicable to the majority of people (considering that the majority may not have a specific educational background but rely on common sense), GPT-4 would be a much better helper?
00:17 🌐 Google Gemini introduces a universal AI model designed to handle various inputs: text, code, audio, images, and video seamlessly from the ground up, a departure from traditional multimodal models stitched together from separate components. 01:51 🌟 Gemini boasts exceptional performance across diverse subject areas, matching or surpassing human expert levels in 50 tested fields, marking a significant leap in AI capability. 02:47 🛠 Google plans to release three versions of Gemini - Ultra for complex tasks, Pro for broad applications, and Nano for on-device efficiency, intending to serve various developer and enterprise needs. 03:16 ⚖ Emphasis on safety and responsibility: Gemini addresses potential harms arising from multimodal capabilities, employing proactive policies and rigorous testing to mitigate offensive or harmful combinations of input modalities. 04:42 🚀 Gemini showcases groundbreaking multimodal capabilities through a demo, highlighting its ability to interpret and engage in diverse conversational scenarios spanning images, text, and interactive prompts effectively. 11:20 📊 Benchmark comparison reveals Gemini Ultra outperforming GPT-4 in seven out of eight categories, establishing itself as the leading large language model in various tasks, including mathematics, coding, and general capabilities. 13:59 🎨 Gemini's reasoning capabilities displayed through a bespoke interface generation for user queries, showcasing its ability to understand user intent and provide tailored experiences beyond simple chat interfaces. 17:26 📚 Gemini's educational potential demonstrated through its ability to assist in learning physics by analyzing and explaining handwritten answers on worksheets, providing personalized explanations and practice problems based on mistakes. 19:33 🧬 Gemini showcases its data extraction prowess from scientific papers, efficiently filtering relevant information from a vast pool of papers, updating datasets, and even generating code for updating graphs, revolutionizing data extraction workflows. 22:03 🍳 A practical demonstration reveals Gemini's integration into everyday tasks, illustrating an AI-assisted cooking interaction that showcases 23:57 📚 Long-context understanding: GEMINI's ability to grasp information throughout lengthy texts up to 32k tokens is a significant leap from prior models, showcasing improved handling of extended content. 25:07 🖼 Multimodal question answering: GEMINI identifies and provides detailed information about specific objects (like plants) demonstrating vast knowledge and potential for diverse use cases. 25:50 🌐 Interleaved image and text generation: Successfully generates blog posts based on instructions, showcasing consistency in generating images related to the provided text,a step up from previous models like GPT-4. 26:30 📊 Chart understanding: GEMINI's ability to extract data from charts and interpret it is a marked advancement, demonstrated by its capability to spot significant data points and present detailed tables. 27:27 ⚽ Video understanding: Effectively analyzes and comments on a soccer video, providing accurate technical feedback on the player's form and technique, showcasing impressive video understanding capabilities. 28:25 🤖 Future implications: Google DeepMind aims to integrate GEMINI with robotics, potentially enabling multimodal interaction with the physical world, emphasizing the exploration of touch and tactile feedback. 29:47 🔮 Future developments: Expectations for significant advancements in future versions of GEMINI, hinting at rapid innovations rather than mere improvements, promising a fascinating year for AI with novel techniques and models.
Considering GPT 5 is coming soon i think they need to push a bit harder. The percentage difference is quite close to the current model and doesn't take into account the other companies that have seperately trained models
The major shift in the next 3 years will be that, as a rule, top level AI labs will not release their best models. I'm certain this has somewhat been the case for OpenAI, Anthropic and Google for the past year. At some point full utilization of a SOTA model will be a strategic advantage for companies themselves to use for their own tactical purposes. The moment any $X of value can be netted from an output/inference run of a model for less than $(X-Y) in costs, where Y represents the marginal labor/maintenance/averaged risk costs for each run's output, no company would ever be advantaged by releasing the model to be used by anyone other than themselves. This closed-source event horizon I imagine will occur sometime in late 2024. There are two reasons why a company may not be inclined to release an extremely capable model: 1. Safety risk: someone uses a model and jailbreaks it in some unexpected way, the risk of misuse is much higher with a more capable model. OpenAI had GPT-4 for 9-10 months before releasing it trying to RHLF and even lobotomized it to being more safe. The Summer 2022 internal version of GPT-4 was, according to Microsoft researchers, more generally capable than the released version (as evidenced by the draw a unicorn test). This needed delay and assumed risks will naturally be much greater with a larger model, both in that larger models, so far, seem harder to simply RHLF into unjailbreakability, and by being more capable, any jailbreak carries more risk, thus the general business level margin of safety will be higher. 2. Sharing/exposing capabilities: Any business wants to maintain a strategic advantage. Releasing a SOTA model will allow a company's competitors to use it, test its capabilities and train models on its outputs. This reality has become more apparent in the past 12 months.
@@Agnes135 I simply do not believe this. There are aspects they could literally just hide away and use for themselves, but then their competitors will get a leading edge. We are talking about the technology space, the most cutthroat part of the economy. By no longer sharing their models they literally lose out on the majority of their value. This is like if two countries went to war, say like Spain versus the Inca's and when they went in they decided not to bring guns and armor because the enemies could get a hold of it and then they would be screwed. Its not like the Inca's could just fire a gun, just like we would not have all the data open source. You also require insane amounts of power for some of these bigger models. If you want to get into security based arguments, google in the past had the ability for a computer to scan an interpret images for over 8 years. They did not release it because they were concerned about Captcha's. Then Open Ai came out and released it without even flinching. Now that someone came in and bit the leg of the beast, only blood and bone will remain.
The last bit of improvement will be based how many users are using it, which will be based on how useful it is to the average person for the tasks they would prefer AI to do. These could likely be based in writing, teaching, research. First step is adoption, second step is optimization.
As soon as I heard them started talking about “offensive” information, my eyes rolled. To put all this work in to something so sophisticated and to then at the same time censor that intelligence to catch itself from sharing certain kinds of information. That’s dangerous and I hate it.
What impressed me the most are 9:26 and 10:32, how the AI is able to identify and process an ambiguous picture and how it's able to process and learn video in real time
The video part makes it obvious how staged it is, it made me really skeptical. The video goes to portrait before he even turns the phone, and when he pauses the progress bar moves before he resumes the video.
Wired: “Only a sliver of Gemini is currently available. Future releases are expected to include multimodal capabilities, where a chatbot processes multiple forms of input and produces outputs in different ways. Just the text-based version has been added to Bard. Gemini is also only available in English, though Google plans to roll out support for other languages soon. As with previous generative AI updates from Google, Gemini is also not available in the European Union-for now.” Guess we’ll have to wait.
*Let's wait for ChatGPT-5 to arrive, then we can discuss the differences between Gemini and ChatGPT. There's no point in beating the drum for Gemini until then.*
When Google first came out in 1998 I remember it was banned at work. lol. I have always loved the simple screen, colors, font, and on-demand information about anything I am curious to learn about. Thank you for keeping the home screen simple with the same font and colors but not liking the word "Bard" too much. I love the word Gemini since that is my horoscope. lol
That's funny. I was too young for that, but I currently use the AI quite a bit in my line of work and there's talk going around saying we're not allowed to be using AI on work computers. My reply: if you think this needs to be banned at work, you're so poorly misinformed that you shouldn't even be allowed to make a decision on the topic. Sounds like your scenario in 1998.
They don't really specify the hardware requirements to run the model, which is really what counts if it is going to be used by consumers. Is it using double the memory of GPT to get 1 or 2% score increase? Even a slight increase in hardware requirements will ruin any chance of it being economically viable for consumer release.
I would add that the central computers which generate the AI systems already use enormous amounts of water for cooling. Apparently 6% of the local water in Iowa, and counting. This of course is a massive problem. Not to mention that the ‘powers-that-be’ are suggesting watermarking screens produced using AI. Let’s see how this all develops.
I remember when chatGPT came out it caught google completely with their pants down. they fumbled a lot in response. good to see that a year later they come out and show they do know their shit.
What I want to know is WHO decides what is unacceptable, what is "responsible", and what will be censored? And how is this censorship going to be exercised?
Chain of learning... Have one untrained model watch and self learn from a primary model. And have a third untrained model learning directly from watching the second one learn from the first... Scale this up as much as you need to... It might not be perfect due to material architecture... But maybe it could highlight the flaws by expounding them...😊
@@morningstarsci Could a person look at, say, the last sensible output, to figure out how that degrades? This interests me, but I'm not intimately knowledgeable about the actual computer science behind the core models... 😔
This is actually a good idea and is kind of the next stage we are aiming for with AI right now. In order to achieve AGI we will most likely have to use a method similar to this. Jeffrey is correct regarding convergence and overfitting but this can be adjusted for, for the most part. Right now AI is built on existing data created by humans and its answers are trained/written by humans, if you think about it, it's pretty much impossible to create AGI using data from humans. Instead the solution would be to simulate data using AI from existing human data to train the next model and then post-train the model using a combination of the previous AIs answers and a human manually picking the best answers from a collection of the previous AIs answers. OpenAI is already experimenting with the first part in regards to simulating data based on existing data and has to some degree implemented the second part using the human feedback and drafts features.
How would it be to have a robot at home that could see and which was embodied with the intelligence of Gemini? It is pretty obvious that Gemini can see and understand the real world. It should be able to navigate the real world well, embodied in a robot.
Kind of... understanding objects and understanding yourself in context of that is very different. Think self driving cars - yes it understands what a road is, but it also needs to understand itself - how big it is, how it moves, how others move, how others might interact with it, how it interacts with the world. This is completely different training and would need to be capable of self awareness to a degree - at least in terms of size, shape, capabilities and interactions when applied into a robot. This is very different than image recognition which is shown and even it was a generic AI - it would need to be aware of itself - think like a human with say a giant backpack bumping into things because you forgot your wearing it.
Maybe, but I think Gemini is on the verge of understanding beyond its training. If so, robots embodied with such intelligence at least in part are going to learn to interact with the world just like humans learn. I am no AI pro. That is simply what makes sense to me.@@goomyman23
Nice to have you covering the AI news so fast. Btw can you check your audio mixing in the future, the video clips you use are extremely loud compared to your voice recording.
Seeing that the model is capable of reasoning and then map the interaction and logic then it generates Flutter codes freaks me out as Software Engineer. REALLY FREAKS me out big time!!! I cant imagine any white collar job that wont be replaced by this AI Model.
A comment above mentioned, it s great, save me a lot of time finding the documentations! But who needs docs when there s no code to write? I m amazed by how short sighted people are
The fact that it needed a - compared to google - small company to start the AI revolution and that Google with all its manpower might have been able to develop such an AI way earlier, but just started after smelling the money and got nervous after ChatGPT, makes me wonder, if they are still as innovative as they claim. I think I'll stick to the OpenAI side for a little longer
If money is what Google after, they would have patterned the Transformer invention (instead of open source it and used by OpenAI) and commercialized LaMDA language model back in 2020.
Of course Google is in it for the money or more accurately, alphabet. To try and defend such a large company that is illegal in itself as it breaks many Monopoly laws is simply silly.
Yeah, they suck at high-level math. I'm assuming it's because they were coded to be language/logic smart, not technical. There could also be a lack of high-level math data, considering the general public doesn't use it.
@@NossDDoss I actually didn't consider the part about public data since that would be the learning basis for AI. It makes sense that the majority of the numbers data be mostly gated for obvious reasons
@@NossDDossopen AI has developed a new AI for math and reasoning/logic. Apparently it is able to compute and solve math problems it hasn’t even been taught yet. It will be interesting to see.
I am a leading authority on education in the AI age. I am a former semiconductor engineer and currently a high school math teacher. I also introduce these technologies to my students. I also educate teachers through TH-cam and other media, including national presentations and lectures at universities.
As a software engineer, i feel we have chopped the branch that was carrying us; I’m now training to become a plumber, hope it’ll some time before robots could diagnose and fix water leaks…
I'm confused - this seems like a multi-media extended press release. WHERE IS THE PRODUCT I CAN ACTUALLY TEST? I'm not interested in AI fanboys telling bme how great their tech is - as a very experienced user I want to test it for myself. Oh, wait: "t's not consumer facing nor available to the public, though thousands of AWS customers including companies like Deloitte and Gilead Sciences can access it for $20 per user a month." I guess they decided on this to limit demand and increase revenue. The video was for sure extrodinary but I'm very sceptical - just too good to be true. Setting the expectation bar waaaay too high. And most of the 'performance is better that ChatGPT-4 were actually quite small differences. It's good that we have competition in the space but this is too much marketing BS.
@@beautifulandtoolate Where did you hear that? My experiences with Bard are consistently not great. But after reading your comment just tested it again using oa few of my fav prompts and it is as poor as it's ever been. Can't see a difference. This prompt results in garbage: "Are you familar with the difference between 'information' and the patterns used to represent it?" ChatGPT is simply better - and by a lot. Stil far from 'excellent' but pretty good.
Great, yes wonderful. I'm sure its amazing. But I don't give a damn if the company doesn't tell me HOW MUCH it will cost and WHEN it will be available. this is like the 3rd time Google has shown us "wow such an amazing product", only for it not to be released yet, or have a release date or price.
I believe with this pace Gemini can easily identify Exoplanets from Kepler project. It could be breakthrough in that project will help the science advance to next level. Add more heuristic algos it will find the life in the universe..! Way to go Google!
😂 language model playing chess. We have seen it in gpt4, plays like 4year old kid that don't know the rules yet even it read all the games in the internet
@@SuliXbrGpt 4 is literally multi modal, Gotham played chess with gpt 4 and it did pretty well and it played moves that stockfish would, it only started hallucinating pretty late into the game.
I engage in various prolific investment by compound interest & leveraging and as well operating with an investment professional Claudia Vecchi Nese, So far I've attained returns of $80k last month
I appreciate Mrs Claudia Vecchi Nese for her good work, I will always share the good news to the world I made it out for the very first time with her, she's truly a God sent to me.
No doubt Mrs Claudia Vecchi Nese Trading services is very good, I invested $20,000 and cashed out $57,700 after 11 days. I still wonder how they get their analysis.
That's a good catch! I wonder if Gemini knew who he was. Anyone with a Google assistant in their house can input demographic information, and the speaker can learn each person's voice in the house so it can cater answers to them. it's entirely possible for it to "know" who it's responding to from previous input. I mean, either that, or they just screwed up on the presentation, LoL.
Will see how well it performs once the masses picks it up. So far gpt 4 has impressed me so much in multilingual communication. I know gpt4 is amazing in responding in Chinese....Not sure if Google is considering this beyond just English....And also how it uses generative voice as a main way of communication.....
It's not like that.. Car & horseman also need a human to drive .. But an AI car doesn't need that human effort.. Painters also can learn photography and do that job .. But in case of AI it will automatically do it... AI will replace the jon culture like currency changed barter system ... @@andysachs83
whats with the lights all pixelated on the begining of the video ? is it the youtube codec that sucks or soemthing ....i never know if this is a problem with my computer or the youtube codec >S anyone ?
OpenAI seems to be sleeping on the job, huh. I hope they announce GPT-5 soon, because Google is not playing around; they want to dominate the global AI market.
Google were always first in AI, but they were very cautious promoting it as a product because ethical and legal concerns. Open AI was done using a lot of foundational research that Google did several years ago.
@@andysachs83 google were not "always first in AI". Sure Deepmind was cool, but LLM usefulness caught them perhaps by surprise. Now they are back to the races IF these promises hold. This is atill very much an open race.
What will happen to our own brain if AI does all the thinking for us? Look at what is now already happening to the quality of youngsters' brains for example who use their 'smart' devices all the time. As is the case with all well-intentioned inventions, also AI must be used with moderation and balance.
@@ollebroms7185 If you have that poor an understanding of what’s to come then you’d do well to dig deeper into the subject than a few posts from Reddit doomer edge lords on the topic.
@@gregw322you should learn economics. The most expensive expense on a company is labor. When everything requires no skills and/or goes million times faster there will be no jobs left but believe in a fairytale that employers will waste their money on labor.
@@ollebroms7185 no one thinks human labor will remain viable. It’s simply that you’re ignorant to the measures in place to deal with these issues. Again, you’d do well to educate yourself on the topic.
Another blabla commercial. And an "example" that's highly customized and cobbled together in a lab is not convincing. Let's see it work in the real world.
@@charlesmendeley9823 And less than 48 hours later, it surfaced that a lot of this stuff was fake, grossly exaggerated, and not realistic at all. Safe to say your salty comment aged like milk, pretty much overnight :-D My mom is fine btw, still happy she didn't raise a sheep ;-)
That's actually pretty good considering current ANN models, that being said as the start of the video indicates it looks like Google is still affirming their core values of Do Evil
Its gonne become sentient way before we know... its learning already just by the one input" predict the next word!" to do that it needs to understand the context of every subject it could possibly be asked about. aswell as when its reading a text or a story it needs to understand the context involving emotions, moral, logic and imagination
AI here is more for entertainment and kid's fun, but what the world needs is a serious knowledge base and a collegue who will guide us in all needed details and to deliver the best in all what we do
It's so relaxed and just solving problems, which is impressive, but I have a feeling most people feel terrified at the same time. This would be a great opening scene to a dark sci-fi movie.
AI has been a game changer in Software Engineering imo. Before sometimes i wasted hours reading stuff, now AI gives me the documentation within seconds.
then next it will learn how you code, improve it, and you will be a man of leisure. Its coming for all tech jobs, hardest and most expensive to fill, and companies can be held to ransom by IT. In the mean time it will "help you"
I work in the tech industery. And let me tell you. By the time a big company announces a new tech, Its actually already been around for at least 5 to 10 years.
I am just a normy who is not informed about tech but i do have enough nerdiness in me. My jaw truly dropped.The first demo felt like a very early version of JARVIS from iron Man. Just imagine we have everything in reality now from a freaking super hero movie. We have the transparent screens, the mansion exists in real life and we now have the smart ai. Its truly unreal how amazing human beings are. These things give me hope so much.
00:01 Google Gemini is a breakthrough in AI with multimodal capabilities
02:14 Google's GEMINI is a family of models with different capabilities.
06:54 Creating a game called Guess the Country with clues provided
09:35 Gemini Ultra has surpassed GPT-4 in nearly everything.
13:52 Gemini's multimodal reasoning capabilities to understand and reason about users' intent
16:00 Gemini's capabilities in creating visually rich and interactive experiences
20:11 Gemini uses Advanced reasoning and multimodal capabilities to filter, extract and update large sets of scientific data
22:05 Google Gemini enables multimodal interactions with AI.
25:20 Google's GEMINI has advanced capabilities for identifying plants and generating text and image content.
27:00 Google Gemini exhibits advanced multimodal capabilities
30:15 Google's GEMINI will bring groundbreaking innovations in AI next year.
thank you
A real one bro saved me from wasting 30 mins
Best comment ever ❤
Google's playlist for all these clips.
th-cam.com/play/PL590L5WQmH8cSyqzo1PwQVUrZYgLcGZcG.html
surpased CHAT GPT? with a staged demo? lol lol lol
Google has had advanced AI for years. Anyone remember Google Duplex making AI assisted phone calls, presented at Google io in 2018? I think they would have integrated it into Google Assistant if people didn't freak out so much. Not that it was wrong to freak out. But now, Chat GPT has proven that many millions of people will incorporate AI into their lives. With all the competition, the Genie is out of the bottle. So, Google is like "alright! integrate everything we've got and hit them with Super Saiyan AI!" It's gonna be a bumpy ride folks!!
🤣
I just cant fathom what 3 months looks like from now. People say 6 months im like things change every DAY
@@chartingwithlivtrue
how long until we don't have to go to work and chill at home because AI does our jobs 1000x better and faster? :)
that is a problem because the current world government diaspora will only value you as a walking target for their new weapon advancement or a subject of radicalization so they have a new reason to go back to start of my reasoning. :P]
The issue is not AI but that humans have no clue what to do with their free time that is constructive and not distructive.
Food for thought.
If you don't need to work and relay on fossil fuels will you still defend their use? //rhetorical question....maybe :P
To me, the significance of Gemini is that it all seems to be in one package. If it can indeed shift seamlessly between modalities, then this is a huge breakthrough and we aren't even in 2024 yet!
the only metric they beat is being the most racist ai model to ever exist
AI is definitely impressive! I have a few points though -
1. Will it not cause decrease in human interaction & dependency?
2. Relying on simple answers like is the omelette cooked, will it not decrease the basic common sense that humans are gifted with?
3. What are the controls that we have in place?
2. imagine the stove just turns off when it sees the food is cooked enough or just stop it from being overcooked and starting a fire
@@Indpendent01 Google could have used this in video instead! This is what I exactly mean, AI is great if used in ways that have a actual positive effect and not replace what we already do.
hot damn, if I ever get bored watching AI news, or it gets redundant, I just wait a couple weeks or so, and its at a whole new level!
Couple of days I reckon 😅
Yes, that's an excellent point. The time between iterations seems to be getting shorter and shorter; if you grew up with SciFi movies from the 80s (like Terminator) and you were living through the 90s / 2000s, you always thought "how is this going to happen by 2030 or something like that, the technology still seems so very far away". And now, just a few years later, it looks like all of what we saw before is "right around the corner".
I can't help but feel excited about our future. 🙂
it’s not prod ready
what makes AI amazing is AI is been used to make AI better which is just insane
thinking the same thing
The good thing about this, is that it’s going to force other companies to release models sooner to the public to compete.
I hear you. Although I feel its the other way around. ChatGPT and others are forcing GOOGLE to release stuff they were gonna wait to release much later.
Yes because deadline pressure is exactly what you want when you're developing a technology that many experts believe could see the end of the human race.
@@benjamininkorea7016 I don’t understand why so many people believe a mass extinction event would be so quick. Haven’t y’all seen The Clone Wars?
Usually, companies only keep the model from the market to test it thoroughly. We will now get models with security risks. That's not a good thing.
@@AustinMulkaMusic The field of AI safety believes that the race condition caused by competition is dangerous. I believe them. The models have emergent properties: when you increase the size of capabilities, they do things we didn't know or intend for them to be able to do. Isn't that at least a little concerning?
It doesn't have to be an extinction event to be bad. The first major application of AI (the social media algorithms) has already nearly torn society apart. It focuses on outraging people since it's the most effective way to harvest their attention, and exposes them to more and more siloed material. People are forced towards extremes, and we have more conflict and division than ever. They weren't programmed to do this; it was an emergent phenomenon. Now facts have become almost meaningless, and everyone lives in their own alternative reality curated by...the AI algorithm.
So...we have clear evidence in the only major application that it can do enormous amounts of damage. It would be very silly not to be worried, and rushing is definitely, definitely not the answer.
Reliability is the question. Google has a history of not doing live demonstrations that were not staged to display the tech abilities. It's going to be interesting to see.
Yeah, I don't believe that this is real.
I thought that way when they released Android :)
I agree this definitely wasn’t a live example 11:15
They are just making public shock, but we have to accept the fact that ChatGPT was a fundamental game chanegr
@@lutaayahuzaifahidris3679 Yes it WAAAAS a game changer, but the NEW MORE COMPLETE GAME NOW is GEMINI !!! ...On neutral basis Google always has had the most robust group of AI Researchers in the world since at least 12 years. Many of those were teachers of current Microsoft researchers.
Very impressive! But they will need to include a voice conversation mode ASAP to compete with open AI. Of course, other companies can use Gemini via API to produce services that will allow you to interact with an avatar that sounds like a real human.
For once the title isn’t click bait, this is actually shocking, I first thought wow then further into the video I thought WOW ! It’s not just keyboard input its video plus voice interaction, amazing !. If they could get this into something like Amazons Astro home robot it would be life changing for old folk or those living on their own etc, someone to talk with all day long, loving it
Until it's in my hands and I can use it to exponentially improve my productivity for a mere $20 per month like GPT4, it's just a hypothetical concept. I could care less about the top speed or handling of a car that isn't even on the market. Mine's plenty fast and I drive it every day. Let's do this again once we can all get our hot little hands on Gemini.
Google has a history of getting people excited for these things just to be silently cancelled. I doubt that’s the case here (because.. duh, it’s about their money), still. You right.
apparently they have already included atleast some of gemini into bard, but i am not 100% sure, thats what they say atleast.
Google is fast becoming a joke
They do have integrated gemini pro in bard and i talked with it. It surely gave better and more detailed responses than its previous version. But the biggest downside for me to drop it was its accuracy. I asked it to give a timeline of all the ai developments that happened recently. It got the events right but not the dates. It said that gpt 4 was released in november 2023 and microsoft bonsai project in 2022. As a student i need reliable answers so bard is yet not suitable for me
“Couldn’t care less”
I was expecting a wider gap in the benchmarks. I Guess we'll have to wait for gpt 5. That Q* breakthrough will have gpt 5 blowing people away.
and Gemini 2.0 :)
And Q* whatever it is.
@@muggzzzzz the gpt 5 is stronger then this it will have Q* capabilities it will be the closest thing we have to ASI
Not sure that GPT5 will have Q* integrated. They are already training GPT5 so for me it's unlikely it will integrate Q*
@@diegoaudinetti7152 thats incorrect it will contain Q*
Love everything except for the potential abuse in censorship for things Google deems "harmful" when it comes to actual discord, or research, especially with things like current events, politics, and hot button issues. Just don't be evil, and stay neutral, unless something is in violation of actual laws.
Completely agree!
Their ethical code was "don't be evil" a while back. They've since removed that and not long after got into the dirty fossil fuel business. The latest which discuss me is the "project Nimbus" that they are involved in, read about it. But don't trust Google, for sure.
amongst the big companies, I trust Google the least to not be biased and try to big tech / big brother spy on people
so I can't say I'm excited for any ai Google cooks up
Exactly. For instance, I noticed it did not incorporate ALL papers, but only the RELEVANT ones (19:48). Relevant to the subject, of course. But what if the system gets programmed to exclude certain studies that contain necessary information on a certain subject, but is deemed "not relevant" because it is not in accordance with a (biased) 'consensus'?
But we automatically assign value to everything we see or hear. We would not accept statements by Biden or Trump as equally true and equally important. No one can hold opposing views on every issue from energy to healthcare.
There are 2 ways to cook chicken in an Instant Pot. One uses 1 cup of water and 10ish minutes high pressure. Another uses 4 cups water and 0 minutes. What would Gemini say if asked what's the best way to cook chicken in an Instant Pot? Combining the recipes would be a mistake. Picking the most popular would suppress novel approaches. What AI can't do is try out both recipes. Would it understand that one is a European method and the other is Asian?
This was my first prompt:
Here’s a puzzle:
two coins are on the table, touching. One is three times the diameter of the other. If I roll the smaller coin around the larger, maintaining contact, how many rotations will the smaller coin make when it has gone thru one orbit?
Of course, Gemini was quite confident of its answer, but nonetheless, incorrect. It answered 5, complete with what looked like pretty convincing arguments and math.
You're explanation/instructions has room for improvements - it was correct in one of 50 different variations where u were not explicit enough in your request.
You should check yourself into a soup kitchen line - Your job has now been replaced by Ai
To solve this puzzle, we can use the concept of the circumference of a circle and the relationship between the diameters of the two coins.
Given information:
- The larger coin has a diameter that is three times the diameter of the smaller coin.
- The smaller coin is rolled around the larger coin, maintaining contact.
Let's assume the diameter of the smaller coin is 'D' and the diameter of the larger coin is '3D'.
When the smaller coin makes one complete orbit around the larger coin, it will make a number of rotations.
The circumference of the larger coin is:
Circumference of the larger coin = π × 3D
The circumference of the smaller coin is:
Circumference of the smaller coin = π × D
To determine the number of rotations of the smaller coin, we need to divide the circumference of the larger coin by the circumference of the smaller coin:
Number of rotations = Circumference of the larger coin / Circumference of the smaller coin
Number of rotations = (π × 3D) / (π × D)
Number of rotations = 3
Therefore, when the smaller coin has gone through one orbit around the larger coin, it will have made 3 rotations.
One thing I love about using GPT-4 via Bing, on my phone, is that it can verbalize the responses. And it does so with a very natural and human like voice in a accurate, concise and helpful way.
Too slow though. Try Voice GPT which is supepr fast but not as smart as Bing. All have some trade offs
@@tranzhuman TOO low?...it's takes just seconds. why does it need to be faster?... how fast does it need to be?.... how fast is sufficient?... so many questions.
@@rw9207Copilot is great but you can't deny that it is slow compared to GPT-4. I put it to the test recently and the slow speed of copilot was unbearable compared to GPT-4
Unsurprisingly, like Chat-GPT, it hallucinates at times e.g. saying that plastics/rubber are less dense than water (in the duck part). The real danger of AI is that we'll struggle to tell what is real and what is fabrication/a distorted version of reality.
Something that will no doubt become common is running the same query in parallel across Gemini, GPT4, Llama, and coming up with a confidence score. We do this sometimes with human experts, like with radiocarbon dating where samples are sent to multiple labs and compared to establish confidence. In an ideal world everything would be checked by 3 humans, but that's impractical -- not so much with models.
@@jwoya Love this concept
Especially politics
Plastic that floats is less dense 😅
@@jwoyaThis is exactly what everyone has to do with everything online especially anything that has been targeted to compete for our attention.
I'm getting tired of doubting everything, because ulterior motives.
This whole talk about safty and filters reminds me a lot of 1984. Concentrate all the power in a single hand and than decided what is "safe" and what is not.
You can always set up your own language system, isolated from the rest of the world and train it the way you want.
How did we ever survive all these years without AI telling us how to bake cupcakes and cheat on our homework? So ground breaking.
That's the whole thing right?, to use AI efficiently, you have to have some brain in the first place. AI enhances smart people but replaces dumb ones. Unfortunately it also prevents smart people from being nurtured in the first place. It creates and solves its own problems,leaving us with a net zero for work efficiency but a negative for society as a whole
We used to read books 🙈🙈🙈
A little bit of thinking, we already gave away with smartphones being essentially a part of us, a part of our brain and daily information gathering. Now we're at a point, where this outsourcing of complex thougts is really, really big.
You have no clue man.. keep fishing and leave AI
I use it to search the internet and not have to deal with all the extra stuff, I also use it to rewrite things, like my resume. Some companies using AI to find new paths in medicine. It will be used to scan people to find disease. It will be used to make things work better. Video games with Ai monsters. Tech will boom.
So this multi-modal method to conception, observation, connection, and response combined with quantum computation which is making regular computers look like a mere calculator now...my mind is blown! I have never felt such exhilaration. Even if this kills me, I have lived to see water being pulled up from a well to this moment, so I would die having witnessed something very profound, unimaginable, and inspiring, but the beauty of it, or the horror of it is that humans created something far bigger than themselves. I used to question the reason for life, but now, I am silenced. This is some kind of proof of something I am too small to process or to which I can put words.
I noticed that GPT-4 has a higher score, specifically in commonsense reasoning for everyday tasks, with 95.3%, compared to 87.8% for Gemini. Does this mean that if we need AI as a tool that is useful and applicable to the majority of people (considering that the majority may not have a specific educational background but rely on common sense), GPT-4 would be a much better helper?
With the lack of common sense in our society right now, any AI will do :)
😂😂😂😂true 👍 😅@@martingosselin6698
Yes, GPT-4 is still the winner so far for the everyday user.
it should be a live demo so we can better assess its capabilities realtime.
They already admitted, that the duck was fake for the most part.
@@mikehotel5652where?
@@mikehotel5652 That's why CNBC says. but don't trust them.
00:17 🌐 Google Gemini introduces a universal AI model designed to handle various inputs: text, code, audio, images, and video seamlessly from the ground up, a departure from traditional multimodal models stitched together from separate components.
01:51 🌟 Gemini boasts exceptional performance across diverse subject areas, matching or surpassing human expert levels in 50 tested fields, marking a significant leap in AI capability.
02:47 🛠 Google plans to release three versions of Gemini - Ultra for complex tasks, Pro for broad applications, and Nano for on-device efficiency, intending to serve various developer and enterprise needs.
03:16 ⚖ Emphasis on safety and responsibility: Gemini addresses potential harms arising from multimodal capabilities, employing proactive policies and rigorous testing to mitigate offensive or harmful combinations of input modalities.
04:42 🚀 Gemini showcases groundbreaking multimodal capabilities through a demo, highlighting its ability to interpret and engage in diverse conversational scenarios spanning images, text, and interactive prompts effectively.
11:20 📊 Benchmark comparison reveals Gemini Ultra outperforming GPT-4 in seven out of eight categories, establishing itself as the leading large language model in various tasks, including mathematics, coding, and general capabilities.
13:59 🎨 Gemini's reasoning capabilities displayed through a bespoke interface generation for user queries, showcasing its ability to understand user intent and provide tailored experiences beyond simple chat interfaces.
17:26 📚 Gemini's educational potential demonstrated through its ability to assist in learning physics by analyzing and explaining handwritten answers on worksheets, providing personalized explanations and practice problems based on mistakes.
19:33 🧬 Gemini showcases its data extraction prowess from scientific papers, efficiently filtering relevant information from a vast pool of papers, updating datasets, and even generating code for updating graphs, revolutionizing data extraction workflows.
22:03 🍳 A practical demonstration reveals Gemini's integration into everyday tasks, illustrating an AI-assisted cooking interaction that showcases
23:57 📚 Long-context understanding: GEMINI's ability to grasp information throughout lengthy texts up to 32k tokens is a significant leap from prior models, showcasing improved handling of extended content.
25:07 🖼 Multimodal question answering: GEMINI identifies and provides detailed information about specific objects (like plants) demonstrating vast knowledge and potential for diverse use cases.
25:50 🌐 Interleaved image and text generation: Successfully generates blog posts based on instructions, showcasing consistency in generating images related to the provided text,a step up from previous models like GPT-4.
26:30 📊 Chart understanding: GEMINI's ability to extract data from charts and interpret it is a marked advancement, demonstrated by its capability to spot significant data points and present detailed tables.
27:27 ⚽ Video understanding: Effectively analyzes and comments on a soccer video, providing accurate technical feedback on the player's form and technique, showcasing impressive video understanding capabilities.
28:25 🤖 Future implications: Google DeepMind aims to integrate GEMINI with robotics, potentially enabling multimodal interaction with the physical world, emphasizing the exploration of touch and tactile feedback.
29:47 🔮 Future developments: Expectations for significant advancements in future versions of GEMINI, hinting at rapid innovations rather than mere improvements, promising a fascinating year for AI with novel techniques and models.
Thank you
cant wait to ask gemini to make these kinds of summaries for vids😂 thanks!
Thanks
I cannot wait to pump my load into a robot.
Thanks bard
Considering GPT 5 is coming soon i think they need to push a bit harder. The percentage difference is quite close to the current model and doesn't take into account the other companies that have seperately trained models
The major shift in the next 3 years will be that, as a rule, top level AI labs will not release their best models. I'm certain this has somewhat been the case for OpenAI, Anthropic and Google for the past year. At some point full utilization of a SOTA model will be a strategic advantage for companies themselves to use for their own tactical purposes. The moment any $X of value can be netted from an output/inference run of a model for less than $(X-Y) in costs, where Y represents the marginal labor/maintenance/averaged risk costs for each run's output, no company would ever be advantaged by releasing the model to be used by anyone other than themselves. This closed-source event horizon I imagine will occur sometime in late 2024.
There are two reasons why a company may not be inclined to release an extremely capable model:
1. Safety risk: someone uses a model and jailbreaks it in some unexpected way, the risk of misuse is much higher with a more capable model. OpenAI had GPT-4 for 9-10 months before releasing it trying to RHLF and even lobotomized it to being more safe. The Summer 2022 internal version of GPT-4 was, according to Microsoft researchers, more generally capable than the released version (as evidenced by the draw a unicorn test). This needed delay and assumed risks will naturally be much greater with a larger model, both in that larger models, so far, seem harder to simply RHLF into unjailbreakability, and by being more capable, any jailbreak carries more risk, thus the general business level margin of safety will be higher.
2. Sharing/exposing capabilities: Any business wants to maintain a strategic advantage. Releasing a SOTA model will allow a company's competitors to use it, test its capabilities and train models on its outputs. This reality has become more apparent in the past 12 months.
@@Agnes135 I simply do not believe this. There are aspects they could literally just hide away and use for themselves, but then their competitors will get a leading edge. We are talking about the technology space, the most cutthroat part of the economy. By no longer sharing their models they literally lose out on the majority of their value. This is like if two countries went to war, say like Spain versus the Inca's and when they went in they decided not to bring guns and armor because the enemies could get a hold of it and then they would be screwed. Its not like the Inca's could just fire a gun, just like we would not have all the data open source. You also require insane amounts of power for some of these bigger models.
If you want to get into security based arguments, google in the past had the ability for a computer to scan an interpret images for over 8 years. They did not release it because they were concerned about Captcha's. Then Open Ai came out and released it without even flinching. Now that someone came in and bit the leg of the beast, only blood and bone will remain.
10:41 I could tell the cat wouldn't jump high enough just by looking at the initial swing. Gemini still has to learn to analyze movement.
The last bit of improvement will be based how many users are using it, which will be based on how useful it is to the average person for the tasks they would prefer AI to do. These could likely be based in writing, teaching, research. First step is adoption, second step is optimization.
Woh woh nhi hai Jo tumhe lag raha hai.
As soon as I heard them started talking about “offensive” information, my eyes rolled. To put all this work in to something so sophisticated and to then at the same time censor that intelligence to catch itself from sharing certain kinds of information. That’s dangerous and I hate it.
Like Communist did, censur things that can hurt their agenda.
@@niclas0jansson Every government does that. The US censored people during the McCarthy era.
I noticed that too! Immediately thought of how creepy and dystopian it sounds. Hal from 2001 would be pleased.
@@niclas0jansson ALL groups of people are guilty of censoring things that hurt their agenda, not just communists. Even the Catholic Church.
What impressed me the most are 9:26 and 10:32, how the AI is able to identify and process an ambiguous picture and how it's able to process and learn video in real time
The video part makes it obvious how staged it is, it made me really skeptical.
The video goes to portrait before he even turns the phone, and when he pauses the progress bar moves before he resumes the video.
Wired: “Only a sliver of Gemini is currently available. Future releases are expected to include multimodal capabilities, where a chatbot processes multiple forms of input and produces outputs in different ways. Just the text-based version has been added to Bard. Gemini is also only available in English, though Google plans to roll out support for other languages soon. As with previous generative AI updates from Google, Gemini is also not available in the European Union-for now.” Guess we’ll have to wait.
*Let's wait for ChatGPT-5 to arrive, then we can discuss the differences between Gemini and ChatGPT. There's no point in beating the drum for Gemini until then.*
When Google first came out in 1998 I remember it was banned at work. lol. I have always loved the simple screen, colors, font, and on-demand information about anything I am curious to learn about. Thank you for keeping the home screen simple with the same font and colors but not liking the word "Bard" too much. I love the word Gemini since that is my horoscope. lol
That's funny. I was too young for that, but I currently use the AI quite a bit in my line of work and there's talk going around saying we're not allowed to be using AI on work computers. My reply: if you think this needs to be banned at work, you're so poorly misinformed that you shouldn't even be allowed to make a decision on the topic. Sounds like your scenario in 1998.
they don’t need ordinary people anymore. Congrats 👏 and sorry for the remaining 8 billion people.
They don't really specify the hardware requirements to run the model, which is really what counts if it is going to be used by consumers. Is it using double the memory of GPT to get 1 or 2% score increase? Even a slight increase in hardware requirements will ruin any chance of it being economically viable for consumer release.
I would add that the central computers which generate the AI systems already use enormous amounts of water for cooling. Apparently 6% of the local water in Iowa, and counting. This of course is a massive problem. Not to mention that the ‘powers-that-be’ are suggesting watermarking screens produced using AI. Let’s see how this all develops.
I remember when chatGPT came out it caught google completely with their pants down. they fumbled a lot in response.
good to see that a year later they come out and show they do know their shit.
It goes from voice and visual prompt in the duck section to ChatGPT style text and code required prompts. Which is it?
What I want to know is WHO decides what is unacceptable, what is "responsible", and what will be censored? And how is this censorship going to be exercised?
not you, not me...that's who... end of discussion... move along now into you cell
Chain of learning...
Have one untrained model watch and self learn from a primary model. And have a third untrained model learning directly from watching the second one learn from the first... Scale this up as much as you need to... It might not be perfect due to material architecture... But maybe it could highlight the flaws by expounding them...😊
It would compound the issues. You would get convergence and overfitting.
@@morningstarsci I guess my question is: Would that be interpretable, or just garbled noise?
@@Robert_McGarry_Poems You would end up with both. The output begins to get non-sensical.
@@morningstarsci Could a person look at, say, the last sensible output, to figure out how that degrades? This interests me, but I'm not intimately knowledgeable about the actual computer science behind the core models... 😔
This is actually a good idea and is kind of the next stage we are aiming for with AI right now. In order to achieve AGI we will most likely have to use a method similar to this. Jeffrey is correct regarding convergence and overfitting but this can be adjusted for, for the most part. Right now AI is built on existing data created by humans and its answers are trained/written by humans, if you think about it, it's pretty much impossible to create AGI using data from humans. Instead the solution would be to simulate data using AI from existing human data to train the next model and then post-train the model using a combination of the previous AIs answers and a human manually picking the best answers from a collection of the previous AIs answers. OpenAI is already experimenting with the first part in regards to simulating data based on existing data and has to some degree implemented the second part using the human feedback and drafts features.
Not holding my breath with Google rolling this out soon...
Never... not at all... not free no google.
Google has already rolled it out. Bard is now powered by Gemini.
@@drawnhere I asked Bard after I watched the video and it said not, but it would be implemented once it was out of development 👍
How would it be to have a robot at home that could see and which was embodied with the intelligence of Gemini? It is pretty obvious that Gemini can see and understand the real world. It should be able to navigate the real world well, embodied in a robot.
Yeah it is really impressive
i agree this what embodied data and robot is the next step
Kind of... understanding objects and understanding yourself in context of that is very different. Think self driving cars - yes it understands what a road is, but it also needs to understand itself - how big it is, how it moves, how others move, how others might interact with it, how it interacts with the world. This is completely different training and would need to be capable of self awareness to a degree - at least in terms of size, shape, capabilities and interactions when applied into a robot. This is very different than image recognition which is shown and even it was a generic AI - it would need to be aware of itself - think like a human with say a giant backpack bumping into things because you forgot your wearing it.
Maybe, but I think Gemini is on the verge of understanding beyond its training. If so, robots embodied with such intelligence at least in part are going to learn to interact with the world just like humans learn. I am no AI pro. That is simply what makes sense to me.@@goomyman23
Imagine a Boston dynamics robot with Gemini, shit lots of jobs will be replaced
22:27 Man if I take a shot every time he says “essentially” I will be drunk before end of the video 😂
Nice to have you covering the AI news so fast.
Btw can you check your audio mixing in the future, the video clips you use are extremely loud compared to your voice recording.
I will check it out
@@TheAiGridPrioritize speed. Audio isn’t the biggest problem in terms of the latest AI news.
This is amazing tech, for sure....at the same time, ya better start planning for big changes in education and in the workplace.
This is very interesting, when can we access it?
Is this released to the public yet?
Seeing that the model is capable of reasoning and then map the interaction and logic then it generates Flutter codes freaks me out as Software Engineer.
REALLY FREAKS me out big time!!! I cant imagine any white collar job that wont be replaced by this AI Model.
A comment above mentioned, it s great, save me a lot of time finding the documentations! But who needs docs when there s no code to write? I m amazed by how short sighted people are
Oh boy the future is going to be wild, trust me.... ...... ... I see it. I . See. It coming. I don't know how can this go right. Really.
While this is an interesting innovation by Google, its also scary what the world will look like in five years from now
The way they make this innocent scares me even more than the AI itself.
So what version of gemini did they use to demonstrate all of these interactive activities?
Presumably there are controls to the controls , so anti malice could be switched off?
Wow, impressive. Gemini, has just disrupted several business models. We wave entered a new age today Wednesday, 06th December 2023
Aka the day before pearl harbor day...
This looks all very very impressive, but when will personal AI assistants be available. That's what everybody is really waiting for?
Soon we will have our God - and we will have made It - with our own Hands.
@@charlesromelus3 Yikes!
How will we afford an AI assistant when none of us have jobs? #PostLaborEconomy
This is Wow!! Is Gemini publicly available yet?
I thought next year
Gemini powers Bard right now.
Don't worry, it will be for only $99,99 per month : ) Technically Free amr? dont let the doors hit you on the way back for more.
¿Cuál es su falla? ¿hay algún botón de desconexión total o de inhabilitación?
You are doing alright, don't let the haters get to you. Keep it up...
Thank you, I will
The fact that it needed a - compared to google - small company to start the AI revolution and that Google with all its manpower might have been able to develop such an AI way earlier, but just started after smelling the money and got nervous after ChatGPT, makes me wonder, if they are still as innovative as they claim. I think I'll stick to the OpenAI side for a little longer
If money is what Google after, they would have patterned the Transformer invention (instead of open source it and used by OpenAI) and commercialized LaMDA language model back in 2020.
Of course Google is in it for the money or more accurately, alphabet. To try and defend such a large company that is illegal in itself as it breaks many Monopoly laws is simply silly.
I'm actually very surprised at the MATH results. You would think this fits (almost) naturally to a computer system
Yeah, they suck at high-level math. I'm assuming it's because they were coded to be language/logic smart, not technical. There could also be a lack of high-level math data, considering the general public doesn't use it.
@@NossDDoss I actually didn't consider the part about public data since that would be the learning basis for AI. It makes sense that the majority of the numbers data be mostly gated for obvious reasons
@@NossDDossopen AI has developed a new AI for math and reasoning/logic. Apparently it is able to compute and solve math problems it hasn’t even been taught yet. It will be interesting to see.
@@caryg4638 thanks for letting me know. I'll look into that.
I am a leading authority on education in the AI age. I am a former semiconductor engineer and currently a high school math teacher.
I also introduce these technologies to my students. I also educate teachers through TH-cam and other media, including national presentations and lectures at universities.
As a software engineer, i feel we have chopped the branch that was carrying us; I’m now training to become a plumber, hope it’ll some time before robots could diagnose and fix water leaks…
I'm confused - this seems like a multi-media extended press release. WHERE IS THE PRODUCT I CAN ACTUALLY TEST? I'm not interested in AI fanboys telling bme how great their tech is - as a very experienced user I want to test it for myself.
Oh, wait:
"t's not consumer facing nor available to the public, though thousands of AWS customers including companies like Deloitte and Gilead Sciences can access it for $20 per user a month."
I guess they decided on this to limit demand and increase revenue.
The video was for sure extrodinary but I'm very sceptical - just too good to be true. Setting the expectation bar waaaay too high.
And most of the 'performance is better that ChatGPT-4 were actually quite small differences.
It's good that we have competition in the space but this is too much marketing BS.
Gemini pro now powers bard
@@beautifulandtoolate Where did you hear that?
My experiences with Bard are consistently not great. But after reading your comment just tested it again using oa few of my fav prompts and it is as poor as it's ever been. Can't see a difference.
This prompt results in garbage:
"Are you familar with the difference between 'information' and the patterns used to represent it?"
ChatGPT is simply better - and by a lot. Stil far from 'excellent' but pretty good.
Here comes the fast tracking of Q*
Great, yes wonderful. I'm sure its amazing. But I don't give a damn if the company doesn't tell me HOW MUCH it will cost and WHEN it will be available. this is like the 3rd time Google has shown us "wow such an amazing product", only for it not to be released yet, or have a release date or price.
Bard has been released several months ago
@@andysachs83Bars is still in experimental mode, has no API, and constantly tells you it can't do a thing.
I believe with this pace Gemini can easily identify Exoplanets from Kepler project. It could be breakthrough in that project will help the science advance to next level. Add more heuristic algos it will find the life in the universe..!
Way to go Google!
This is the best video I have seen on Gemini. Now I have a clear understanding of Gemini and its future.
I wanna see gemini playing chess against stockfish
😂 language model playing chess. We have seen it in gpt4, plays like 4year old kid that don't know the rules yet even it read all the games in the internet
@@Xizario2 gpt4 is not multimodal it does not understand chess. I want to see if gemini being multi modal makes it understand chess.
@@SuliXbrGpt 4 is literally multi modal, Gotham played chess with gpt 4 and it did pretty well and it played moves that stockfish would, it only started hallucinating pretty late into the game.
Thank you Lord Jesus for the gift of life and blessings to me and my family $32,120.47 weekly profit Our lord Jesus have lifted up my Life!!!🙏❤️❤️
I engage in various prolific investment by compound interest & leveraging and as well operating with an investment professional Claudia Vecchi Nese, So far I've attained returns of $80k last month
😱Sounds familiar, I have heard her name on several occasions.. and both her success stories in the wall Street journal!
The economic hardship, recession, unemployment and the loss of job caused by covid pandemic is enough to push people into financial ventures.
I appreciate Mrs Claudia Vecchi Nese for her good work, I will always share the good news to the world I made it out for the very first time with her, she's truly a God sent to me.
No doubt Mrs Claudia Vecchi Nese Trading services is very good, I invested $20,000 and cashed out $57,700 after 11 days. I still wonder how they get their analysis.
Gemini failed to ask the daughter’s age. Seems hard to plan a birthday party without this basic info.
That's a good catch! I wonder if Gemini knew who he was. Anyone with a Google assistant in their house can input demographic information, and the speaker can learn each person's voice in the house so it can cater answers to them. it's entirely possible for it to "know" who it's responding to from previous input.
I mean, either that, or they just screwed up on the presentation, LoL.
It already knew that. Lol
Will see how well it performs once the masses picks it up. So far gpt 4 has impressed me so much in multilingual communication. I know gpt4 is amazing in responding in Chinese....Not sure if Google is considering this beyond just English....And also how it uses generative voice as a main way of communication.....
Of course, the true test of an A.I. is whether it can make a fun and engaging (but still casual) commander deck for EDH. That needs a benchmark too!
What’s truly amazing is how many jobs, careers, and lives this will destroy.
Or evolve
@@googletropcurieux8670nope...destroy
This is exactly how it was when painters saw the photography, when horsemen saw the cars and when solar power was born. It's an evolution.
It's not like that.. Car & horseman also need a human to drive .. But an AI car doesn't need that human effort.. Painters also can learn photography and do that job .. But in case of AI it will automatically do it... AI will replace the jon culture like currency changed barter system ...
@@andysachs83
Why not say how many jobs, careers and lives it will build?
Now the real war begins 🎉
I thought this is one of the episodes of Black Mirror.
“I hope you’re excited about Gemini’s capabilities”
I’m not because I’m a web developer and out the fucking job😭
whats with the lights all pixelated on the begining of the video ? is it the youtube codec that sucks or soemthing ....i never know if this is a problem with my computer or the youtube codec >S anyone ?
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, It's facinating to see how advanced Google's Gemini is handling multimodal interactions!
Bom dia!
Put...q..par... é a Skynet essa por....
OpenAI seems to be sleeping on the job, huh. I hope they announce GPT-5 soon, because Google is not playing around; they want to dominate the global AI market.
Shut up, open ai is definitely winning AGI race.
@@krishanSharma.69.69f You got stock in Open AI bro?
@@krishanSharma.69.69f that is not something that anyone can know at this moment.
Google were always first in AI, but they were very cautious promoting it as a product because ethical and legal concerns. Open AI was done using a lot of foundational research that Google did several years ago.
@@andysachs83 google were not "always first in AI". Sure Deepmind was cool, but LLM usefulness caught them perhaps by surprise. Now they are back to the races IF these promises hold. This is atill very much an open race.
What will happen to our own brain if AI does all the thinking for us? Look at what is now already happening to the quality of youngsters' brains for example who use their 'smart' devices all the time. As is the case with all well-intentioned inventions, also AI must be used with moderation and balance.
It’s to take over and eventually replace us.
Gemini is updated version of Library Management System we develop , am I right ?
It's impressive but those models are likely to make a lot of people redundant soon, I hope they will have a good answer for that issue if it happens
this is SO exciting!
It's exciting to soon be unemployed and replaced by AI. I'm looking forward to it. Poverty here we come!
@@ollebroms7185lol
@@ollebroms7185 If you have that poor an understanding of what’s to come then you’d do well to dig deeper into the subject than a few posts from Reddit doomer edge lords on the topic.
@@gregw322you should learn economics. The most expensive expense on a company is labor. When everything requires no skills and/or goes million times faster there will be no jobs left but believe in a fairytale that employers will waste their money on labor.
@@ollebroms7185 no one thinks human labor will remain viable. It’s simply that you’re ignorant to the measures in place to deal with these issues. Again, you’d do well to educate yourself on the topic.
Lol this has not aged well
I filtered the comments to newest first to find you're comment, because I was about to say the same thing. lol
You just know that Gemini wont be able to tell us what a woman is.
I tried my gemini to help my child with math paper but it says it is not allowed to access camera? Are these capabilities not yet released?
This is a great video that explains it all in one place. On another note, the vocal fry at 19:52 makes me want to flip a table 😂
Yo, it's the most racist ai
Probably the real natures are truth!
@@janesung819 its literally programmed this way
What the quack
Another blabla commercial. And an "example" that's highly customized and cobbled together in a lab is not convincing. Let's see it work in the real world.
Stop complaining. I feel sorry for your mom.
@@charlesmendeley9823 And less than 48 hours later, it surfaced that a lot of this stuff was fake, grossly exaggerated, and not realistic at all. Safe to say your salty comment aged like milk, pretty much overnight :-D My mom is fine btw, still happy she didn't raise a sheep ;-)
Stop using epic titles for mediocre projects
Jebaited😂
If this is a mediocre project, please show me an epic one.
OpenAI just entered the chat. 🔥great content as always!
Interesting video with a big flaw. The soundlevels vary a lot betveen speakers. Maybe a job for Gemeni to adjust soundlevels?
That's actually pretty good considering current ANN models, that being said as the start of the video indicates it looks like Google is still affirming their core values of Do Evil
Its gonne become sentient way before we know... its learning already just by the one input" predict the next word!" to do that it needs to understand the context of every subject it could possibly be asked about. aswell as when its reading a text or a story it needs to understand the context involving emotions, moral, logic and imagination
AI here is more for entertainment and kid's fun, but what the world needs is a serious knowledge base and a collegue who will guide us in all needed details and to deliver the best in all what we do
how does it know when to wait for a request from you or go ahead and make an assumption itself?
It's so relaxed and just solving problems, which is impressive, but I have a feeling most people feel terrified at the same time. This would be a great opening scene to a dark sci-fi movie.
How to use free Gemini in Canada? We are ot even able to use google bard here. Does Gemini superseded the bard?
Google hasn't been trustworthy for at least a few years now and have gotten too large and powerful to care.
and if you dig deep into its past, some of its original seed money came from a cia shell company...
AI has been a game changer in Software Engineering imo.
Before sometimes i wasted hours reading stuff, now AI gives me the documentation within seconds.
then next it will learn how you code, improve it, and you will be a man of leisure. Its coming for all tech jobs, hardest and most expensive to fill, and companies can be held to ransom by IT. In the mean time it will "help you"
Google Gemini just verified the GIGO rule: Garbage In, Garbage Out
This technology in the apple vr will make us immediately a more advanced species
wake up...please
I work in the tech industery. And let me tell you. By the time a big company announces a new tech, Its actually already been around for at least 5 to 10 years.
I am just a normy who is not informed about tech but i do have enough nerdiness in me. My jaw truly dropped.The first demo felt like a very early version of JARVIS from iron Man. Just imagine we have everything in reality now from a freaking super hero movie. We have the transparent screens, the mansion exists in real life and we now have the smart ai. Its truly unreal how amazing human beings are. These things give me hope so much.
is the homework example done by Bard ?
OMG, I love the details 😮 I love your stait forward analysis 🧐