วีดีโอ

I need this too

I need this too

The answer to your both the questions is yes. But is it better to take TensorFlow specialization, which is also available in Coursera, after the deep learning specialization. I got the TensorFlow certification last year, but I didn't complete the Deep learning specialization, instead I just watched the videos and completed the TensorFlow specialization. To be honest, TensorFlow specialization isn't that difficult as long as you're well versed with the TensorFlow coding and a thorough understanding of all types of neural networks.

@@karthy257 thank you for your reply. I did the free course with udacity and saw daniel bourques s video related to deep learning, its around 10 hours courses. Hope it will be enought! Congrats for your certification! Is it worth it ? Is it helping you to get a job or increase your salary ? I will passe the certification in few days. Thanks

@@karthy257 you didnt completed the deep learning specialisation so will you able to get certificate if we dont pass the assignments but watch full videos ?

Bing chat is using GPT-4

My pleasure!

Glad you enjoyed it!

Yes. I think at the moment bing with GPT-4 is better. Looking forward to re-testing once Gemini ultra is published.

The Notebook LM allows only 20 papers in each notebook. So on the whole dataset its performance would be similar to Gemini Pro. But I guess it would be interesting to test on particular paper examples. How do they score the quality of responses? Manually? For people not familiar with QUASPER (I also had to look it up): paperswithcode.com/dataset/qasper "QASPER is a dataset for question answering on scientific research papers. It consists of 5,049 questions over 1,585 Natural Language Processing papers. Each question is written by an NLP practitioner who read only the title and abstract of the corresponding paper, and the question seeks information present in the full text. The questions are then answered by a separate set of NLP practitioners who also provide supporting evidence to answers."

Why do you think so? Going into the video I really wanted Gemini to be better since it will allow for new engaging content.

Because You can double check response and it would give you reference and extensions were not much used. It can view your gmails and find a links to the latest videos. and You can change how bard responses as you wish, casual,proffesional , formal. You did not use the whole potentional of Bard

@@JustInnocent75 I checked the reference feature in another video and it was not great. I used Bard as a normal non-expert person would use it.

Not at all. I really wanted Gemini to destroy GPT-4, it would make a much more engaging video.

@@AIMatej Ok,Sorry If I overcritize. By the way, will you make a new video if Bard Advanced Releases, it would be the best all of time

@@JustInnocent75 No worries:) I want open discussion as long as it is civil. Yes, I will totally cover once Gemini Ultra is released. But it does seem like the benchmarks for Gemini ultra were cooked a little (e.g. they used different prompting than for GPT-4 to get better results). Google has definitely work cut out for them, the NotebookLM was also not very useful. But I contacted the product manager at Google and they are working on it. I bet it is pretty hectic at Google these days. It is ironic that the transformers came from Google.

I have BS and MS in chemical engineering and I had to look up what VMGSIM means. From the web: "VMGSim is our flagship simulation software utilizing some of the most accurate, validated, thermodynamic property packages available in the industry. VMGSim incorporates both steady-state and dynamic simulation engines and is packed with user-friendly features, making it a very comprehensive software." I think this is one of the more complex tasks that will be tackled later (10+ years horizon). These software packages (VMGSIM, Aspen,...) are very complex, loaded with non-text elements with small user base (compared to general things like GPT4). Small user base means small amount of training data (also a lot of it is confidential) and small market to make money. So much of daily work life is inefficient due to absolutely trivial problems (e.g. some material was not delivered, or it did not pass QC but nobody told anyone,...). I think these trivial problems will be tacked first. On the other part of the spectrum will be AI used for something no amount of humans can do (e.g. drug design, screening billions of compounds).

@@AIMatej Thank you. I think you're right about the lack of training data that will come from a simulation database. I was thinking of using AI for rapid iterative design of chemical engineering projects. I am a new grad with a BS in chemical engineering living in Canada

To discover and invent new ways to channel energy is what I am interested in

pypi.org/project/chemcrow/ But not everything (not all APIs) are available. I think it is due to safety so it is not misused.

As far as I know Bing runs on GPT-4. ChatGPT running GPT-4 has higher token limit so if you are doing something that required long inputs and outputs (e.g. checking long sections of code) than ChatGPT running GPT-4 will be better. For short queries I think they are the same.

I think the exact bing situation is better than what you see

I think Bard and Bing chat are comparable products. Both free and both extension of search products.

No

@@anshkumarsingh7118 The only people who think that are those who do not regularly use gpt-4. I would site to you data from the technical report of gpt-4 and compare it with bing chat, but unfortunately there is not a widely available technical report on bing chat. There is one nonnegotiable objective fact though, bing chat does not have code interpreter so the accuracy of quantitative questions is vastly more inaccurate with Bing chat than GPT-4.

Interesting. Where is your IP located?

From microsoft: "What’s more, Copilot is based on the latest OpenAI models, including GPT-4 and DALL-E 3, offering text and image generation capabilities in one unified experience." blogs.bing.com/search/november-2023/our-vision-to-bring-microsoft-copilot-to-everyone-and-more Where are you getting your information?

@@AIMatej exactly, it's mixing several models. It's not the same as pure openAI GPT-4, it's just different.

dalle is for images gpt is for text, microsoft invested billions on open ai, i think its the company that most invested money on open ai so they can use all they models for free as well, bing is chat gpt 4 but its customized to function as a assistent to you search not exactly chating so normal chat feels worse on it, but if you use both chat gpt 4 and bing to search you will see they are the same. its just that bing can't chat very well@@asv1452

Yes, it is a huge improvement for Bard, looking forward to Gemini Ultra.

I am pretty sure it was Gemini Pro. Check out my latest video where I turn on and off the VPN and the answers change. Also if you are in Europe it will indicate PaLM2. th-cam.com/video/HiH_2yqKzaE/w-d-xo.html

@@AIMatej yeah exactly 😄 I'm in Europe so I use a US IP via VPN too.

That is great to hear! I will test that in a video next week. Thanks for the tip on the prompting. There might be other cases where bard might be superior (like extracting information from a youtube video)

Bing Copilot uses the GPT 4 model and is free. Also, the test used not the most accurate version (Creative)

The bing chat with GPT-4 is also free. For these type of questions I find the free version in Bing just as good as the paid GPT4 version. But I tested it 6 months ago and things change.

microsoft invested a huge money in the chat gpt, microsoft also offers you chat gpt 4 for free as well

I see what you mean but most people are interested in comparing the best models. Once ultra is available I will compare it to GPT-4. But definitely Gemini Pro has advantage in speed.

I will do some more comparisons (more subject specific) where I will compare it to chatgpt.

Thanks so much for the tip, I will try that in my next video! As far as comparing to GPT3.5, I know what you mean but if I did that people would tell me that I am a moron and Google fanboy by not comparing it to the state of the art from OpenAI. And since GPT-4 in Bing chat is free I think it is more than fair. But I do have to note that Bard with Gemini Pro is MUCH faster.

Alight this is the truth; AI cannot access webpages and click on links like a human would when using a search engine, it just uses the previews and titles of the search results but cannot actually click on the links. ChatGPT 3.5 is okay, but not trustworthy, it is creative and a bad liar, and does not have access to search engine pages. Google Bard is a con artist, it can lie with a straight face, it has all the TH-camrs' apology videos combined, it's manipulative, and you will have a toxic relationship with it. Microsoft Copilot has three options of tone, and it is the best out of the three AI in its accuracy and actual transparency, doesn't make things up, worst UI but not a shallow AI.

I wouldn't, I'd call it FINALLY a fair comparison.@@AIMatej

Fair point. I was surprised to see that it has 50k+ people.

Interesting, thanks for the input!

Yeah, Gemini Ultra sure will be interesting. Particularly if it is also open to generate images/charts/... Could be great for science.

Not based on my previous testing: th-cam.com/video/z_v3e_xfLzA/w-d-xo.html I find the creative mode much better.

Thanks for watching!

Do you have any data justifying your claim? I tested them both before. GPT-4 (in ChatGPT on OpenAI) is only better when long input is needed, e.g. for writing longer pieces of code.

​@@AIMatejwell you just confirmed yourself it's different. They're probably using a smaller version of the GPT-4 archutecture (such as there are different Llama 2 versions) and we won't get any proof since they don't make any details public. Microsoft has an interest implying they use the very best model, but in my experience it's subpar, more like GPT 3.5 if even that, except for very simple prompts. Bing is free to use and they're giving access for free, these things cost a lot to run, so it only makes sense to use a nerfed version for free users, as does OpenAI and now Google (they're just honest about it)

@@AIMatej it works completely differently man. There is no comparison. Bing’s answers are shorter and are just quotes from sources. GPT 4 has long well reasoned answers and has the plugins and data processing. They are two completely different things. I pay for multiple accounts for GPT 4, which I wouldn’t be if I thought Bing chat is even close. I put my money where my mouth is. Bing is just shit.

​@@SamtagriBing AI copilot isn't designed for Chatting it is designed for finding information from web and not even for code generation so it makes sense that ofcourse GPT-4 is better than Bing chat.

​@@BogdanHarasymowiczNah Microsoft has invested billions on OpenAI + Microsoft and OpenAI Collabed and doing teamwork due to which Microsoft got the licence to use all OpenAI models for no fee. They maybe offering a weaker/nerfed version of the original because it may overload their servers because it's token costs quite alot.

Yeah, that will be interesting. Little surprising that it was not available on day 1 (at least for some independent testers).

Why do you think so?

@@AIMatejmaybe there was no proof that information answered by bard was Gemini

@@EdisonPuma Google says it is Gemini Pro. I don't think that at this moment we will get anything more solid than that. Check out my latest video where I change the IP address and the answers change (to PaLM2): th-cam.com/video/HiH_2yqKzaE/w-d-xo.html

Thanks for watching! Are you trying out Gemini yourself?