Wow imagine having a small gadget in your pocket with which you can order food, play music, voice your opinions to the world for some reason, and that also wakes you up in the morning! I can't even imagine such a gadget, it sounds like absolute magic to me.
You mean "chat, how long is the line at the soup kitchen and is the bread line out of bread?“ "Comrade, dwell not on the selfishness of one's own hunger, but on the hunger of thy bothers. Remain hungry that they may eat."
What the general public doesn't realize is how much of a breakthrough it was to make an MLP actually produce a text sequence that took context into consideration and produced a response consistent with it. An nobody though it was possible because OpenAI literally had to have access to entire datacenters worth of data an training to do it. From a technical point of view, LLM's are some of the most exciting advances into pattern matching and time function predictors that we had in AI for maybe 3 decades. But this means jack squat in the real world. There is so much more complexity and context that humans and even animals need to process in order to make meaningful actions over the world that AI isn't even close to matching it. AI still needs to solve countless problems in order to even begin to approximate human reasoning: associative memory, higher function abstractions, body-sensory reactivity, self-regulated pain-reward cycles, and many more aspects of what makes a "human" response human. We have trained a machine to repeat a bunch of sequences, and we are ascribing personality and intention to those repetitions in the same way we curse the gods when the roulette wheel draws black for the 10th time in a row. We need to make much clearer what we want out of AI: -If we just want a autocomplete on steroids then we need to stop talking about AI and expecting the AI to understand concepts or to know what "everyone" means. -If we want the AI to resemble a human response then we need nothing less than to create artificial life, because without a body and self-intention the AI has no base ground for interpreting sensory data, so its not unexpected that it will simply spit back nothing more that what you put in. But giving AI a "body"... that is a whole other can of worms.
Yes I'm sure you're right and all the tech giants that are collectively investing trillions into AI infrastructure and adoption have no clue what they're doing
We aimed for centaurs: human brain + horse power for execution. We got the inverse-centaur: horse brain taking decisions + humans running after it to fix its mistakes...
openai, who have produced the most groundbreaking models every time, have described GPT4 as an "orca" and GPT5 as a "blue whale". so when GPT5 comes we'll hopefully get one last experience of that feeling of wonder, seeing it produce some unbelievable behavior, before we get used to it and realize how predictable and robotic it is.
@@tacokoneko I wish to agree with you. But unless the model/training changes fundamentally we still going to have the same errors, they will just be harder to spot because they will sound more convincing. They can improve the LLM's vocabulary 1000x, but no matter how fancy, LLM will still speak nonsense. In the end, it is still linear algebra and probabilistic distribution, even with all the scripts, Indians, and other things that they put on top to make us believe that the chat is actually reasoning.
WHO THE FUCK WOULD LET AN AI ORDER THEM FOOD??? I will never not want to select what I want from a menu with photos and confirm everything is correct before ordering. It's not like it takes that long to do it from your phone...
Imagine when an update creates a defect and the AI orders something that the user is allergic too... Update: Version 4.1 fixes small error where we accidentally unlived a few users with nutrition sensitivities.
I run selenium RPA scripts professionally on external applications via selenium grid and, let me tell you, if you don't yourself own the external application and have control over its development, keeping these runtimes in sync is a stateful, unpredictable, hard-to-maintain mess. The UI becomes the system of record, but it's in no way ACID compliant and rollbacks are more or less impossible. Always call an API if you can.
Hey do you know if the software(s) inside this thing could run on a normal Android phone? Or if it requires some kind of special chip that isn't on Android phones? Because I have a feeling they could've just write an Android application that would run the same exact code... The fact it is its own device is a good pretext for the price tag while almost nobody would pay even half the price of this device for the same thing but as an Android application, even if it was executing the same exact code.
We aimed for centaurs: human brain + horse power for execution. We got the inverse-centaur: horse brain taking decisions + humans running after it to fix its mistakes..
@@Oglokoog if i understand how playwright works the web documents are not fed into the AI. Playwright javascript is rather injected into the website to query document elements and trigger virtual clicks. The main problem: hard coded scripts. When they first announced this i thought if this was true LAM it would involve some kind of image recognision and kernel level human input simulation that will work on any app or interface in foreground.
I would like to take a small amount of credit for “why is Eva everywhere”. I don’t know what I’ve unleashed on the industry but I’m proud as hell of her
At this point, I've seen so many scams and useless products that are downright scams, I've developed an intuition for spotting it. The issue is that I don't feel particularly offended and just ignore it and move on. But perhaps I should try to stop them and look into the things to warn others. Maybe that is my personal scam: The scam we made along the way.
I just sprinkled AI everywhere to get normal VC funding for my company which has a legitimate product (without AI) and it worked. VCs are like AI, they just do pattern matching, they don't really understand shit.
Prime is modest. If he has a Masters in AI, he knows more about AI than Nvidias CEO, who is nothing but an Electircal Engineer and mainly business man who hires people to study AI. Most people who study AI, literally just study LLM's and programing hoping to accidentally create an AI. It's not as cool as people think. Like Prime said, it's not magic. The Rabbit shit is a scam and I hope that someone finds a way to take all 20million they made away from them. Nothing worse than manipulator preying on the people.
It's so damn frustrating for those of us in the general field, too. Nothing against LLM and ML people, but that's only a small fraction of the whole field, and places where there is really cool cutting edge ideas going on are just being ignored because they're not LLMs.
I dont think its right to discredit Jensen Huang because he seems to lack academic credentials. There are literal Turing Award winners with only a Bachelor’s degree.
It was one Christmas and I bought EA game Jedi knight something. It took me 2 hours and my sanity to finally pass their 21 step captcha. It was insane.
App navigation could be fully automated if there was something like OpenAPI contract specification for UI stuff. Also OpenAPI(swagger) could actually make it possible to interact with that many apps.
All future rabbits: User: order me breakfast. Rabbit: Our Doordash rabbit is down for maintenance, instead we're connecting you with the Cloudflare Trust and Safety team.
circa 27:00 "To start off, I would love to know more about your background in tech.." That butt-hurt intro right there, negates any credibility of the remainder of their response.
Sending the html and letting a modern LLM figure it out is absolutely hopeless for the firseeable future. That would require some understanding, and searching the internet for some website with a similar collection of words wouldn't work.
Peter Molyneux was (big emphasis on was) a great game designer for his time and very influential in some genres even up to today. You could even put some of his bullshit down to getting overexcited and thinking interviews were brainstorming sessions. Most "AI people" making noise aren't even involved in training models beyond maybe a LoRA or two and merges. I have a lot of problems with the baked beans salesman who lucked and half conned his way into software but for once I think the comparison is unfair to him.
You could train an LLM against both html and a page screenshot to generate those playwright/selenium code. Inject changes into dataset, see how it adapts. I think it's very much possible and actually gonna happen at some point in near future, for something like that the tech is not the limit, maybe the processing power to train the thing is.
Jesse Lyu's background is in finance and marketing, yet says only those with some sort of valid AI credentials could possibly understand his gadget. AGI is just 5 months away.
What's most strange about this is that while integration is hard, some services are obviously get more used than others. This means implementing Uber, Amazon, and other big/huge names directly (API) produces more stable and faster replies. They didn't even do that.
These dudes saying their LAM is real reminds me of Nikola Motors claiming their hydrogen tractor was real truck, and not just something they pushed down a hill to make it look like it drives.
Your TH-cam channel is the best thing I have seen on the internet in so so long. Thank you for saying what my brain says every day at work. Ai (the bullshit bot variety aka llm) is a terrible solution to all of the problems in this video. 15 years software engineer. You complete me.
6:42 Well, as a Selenium enthusiast myself, I would actually love to have AI-accessible pages. Makes web-scraping orders of magnitude reliable. More bots to the god of bots!
This is literally like something I created for a contractor to handle mass call volumes - use several nlp models to detect an intent from a given context, trigger an AWS lambda function to call a script and do a predetermined thing. It was utilitarian but NOT in any way "intelligent".
Fridges are a peculiar tools. The smaller it is, the more things you'll manage to fit in them, and the bigger it is, the more you feel inclined to leave empty space, and pretend it's full when it's only half full.
Instead of paying for an AI Bot, why not hire a real person, a Virtual Assistant that is more than capable of helping you. There tons of people from 3rd Worlds willing to do that job since US Dollar is superior compared to their money. I know coz I'm one of them :3
The random locations when replying to "near me" could just be it repeating responses from the traning data. You search on google restaurants near me, google correctly replies with restaurants near you in texas, rabbit learns to say texas in reply to "near me". AI is great.
I worked with some Asia based techpreneurs and "xxx is a marketing term" is the mantra over here. It's very typical to make a fully automated micro showroom as the shop front and then fully manual production room with hundreds of underpaid employees which would never ever be shown to the customer. It's called good business.
Last comment, Prime i found you this year. Been coding since 1994 when i was 9, and you brought back enjoyment and passion to it this year for me. Appreciate your vids.
25:51 - "We welcome you to invite experts with valid certificates from the field to further look into this" :: "and we would very much appreciate it, because we don't have any of them ourselves, all we have is a fleet of Playwright automation engineers."
The fact that we configure language models by setting out a list of rules in natural english and telling them to follow it is simply astonishing. These things are crazy smart.
Last year I thought machine learning was what I wanted to do, then I learned ACTUAL machine learning engineers were getting shunned for the generative AI prompt engineering crowd and all the promises that came with this new wave . Actual machine learning is pretty boring , you spend months on a single project that may not see the light of day . The cool stuff is developing whole integration of AI through APIs
@@ci6516 I bet all of those AI enthusiasts would fall asleep the second they get lectured on the core that makes AI even a thing. It's pretty much like with every new trending technology: People don't even know the basics of it, and yet make conclusions. Not to say I'm the guy who knows about it all, but at least I'm not hiding the fact that I have no idea what I'm talking about.
I'm pretty sure in the future SEO will not be a thing. Instead we're going to have ASEYO (Artificial Search Engine Yield Optimization) and it's going to be named ASEYO because it sounds like a korean hello and korean dev will create it.
The other video Coffeezilla did on this topic shows how the Rabbit people were working on an NFT product until the market crashed, and then they quickly pivoted to Rabbit.
33:16 I wonder if they don't have a GPS on this thing because it would take too much power. They use the Internet for localisation. For instance on a vpn from home, the work laptop thinks it's on a different city.
18:16 😂 I laughed so hard when u blamed Tailwind for class overload hell in HTML, that can't fit in AI token. Yes, tailwind classes are useful, but it's a pain trying to create varied animated ui elements. It's probably just easier to create 1 scss mixin function with multiple properties, that can be reused. I have a love/hate relationship with Tailwind atm 😂
Bro I just got a click up ad on this video and the WHOLE ENTIRE APP is "things that suck about jira" wtf is this lol. Not a single mention a clickUp feature just "jira sucks and we do it better" HOW DO YOU DO IT BETTER? Like, I hate Jira but you told me nothing about YOUR product
The funny thing is they are so stupid in a real world application and as a developer myself if i were to make this concept it would have been so easy using chatgpt but they messed up going the playwright route instead of working with companies like doordash and implementing a Application interface allowing the Language model to turn words into actions.
Multics actually was created and Wikipedia says Multics 12.8 was released August 10, 2023. UNIX came to be because they felt Multics was taking too long and the hardware requirements were too high. (i.e. They felt it was over-engineered)
rabbit choose to be a scam, that’s their call. but don’t call llm powered app chatgpt-wrapper. it’s like calling backend database-wrapper and web app html-in-a-specific-order
That's exactly why my initial thought when I saw this device was "AI navigating through UI, hell no. What a waste of resources. Create an API platform...". UI navigation is such a slow and unreliable process...
I remember that their ads were everywhere on Facebook. Funnily enough, no one knows what the product was or what it was supposed to do. They never answer those questions.
The problem is that less than 1% of people who are currently trying to be a programmer actually loves it and want to be a master of it, so the junior or even mid-level market is still going to be terrible for the foreseeable future.
As someone who uses Playwright professionally, I was unaware it was actually AI. Resume updated.
these days AI is everything and everything is AI
"Master AI Developer"
Is playwright just selenium with a new name?
@@katanasteelsorta, cypress is the new selenium and playwright is the new cypress. Playwright is very good imo
Playwright is used for test automation. It’s great, but breaks often and needs to managed over time.
14:42 someone in chat saying "plain text is encryption if you can't read" is the best thing I've seen in a livestream chat for some time 😅
Damn
Theres like hundreds of "plain text" languages i can't read.
I used to think it was a skill issue now i see its just encryption.
Every time I see the chat in one of his videos I’m dumbfounded at the remedial shit they actually decide to say. 😅
DeepL best bruteforce decryption tool lol
Base64 is encryption confirmed
Gatekeeping v1.0 - “You just don’t understand blockchain”
Gatekeeping v2.0 - “You just don’t understand AI”
Both have their use cases, including scams.
Scamming, not gatekeeping. Nobody said you can't make or use those, but there are a lot of people who are using them as buzzwords to turn a profit.
Gatekeeping v2.0 - “You just don’t understand NFT”
Gatekeeping v3.0 - “You just don’t understand AI”
AI
Actually Indians
💯
Actually Playwright
API: Actually Playwright & Indians
LAM
Lotsa Asian Men
Alotta Indians
You mentioned "it feels like we're in NFT land", and that's funny because Rabbit used to be an NFT company.
AI hype taking off so close after NFTs died made a lot of the NFT people I knew hop over to the latest easy money.
You can tell Prime's giving them the beneift of the doubt.. The fraud background of these guys speaks enough to their intentions
Wow imagine having a small gadget in your pocket with which you can order food, play music, voice your opinions to the world for some reason, and that also wakes you up in the morning! I can't even imagine such a gadget, it sounds like absolute magic to me.
after you put it like that, yeah. phones. and they did not made it an app. crazy
If only such a device existed.
it is literally just an app on an android device. they just hid this fact to make it seem like something novel
and what if you could communicate with it using your voice, through e.g. something we could call “Siri”? 😆
The Rabbit CEO has resting scammer face.
Weirdly... I know exactly what you mean.
His face gives me "learning dissability" vibes, so I'm very confused how anyone gave him money.
He's Jesse Lyu
Like Vitalik Buterin. 😔
"Chat order for me 2 pack of milk, if they have eggs order 12."
"The grocery have eggs, Ordering 12 packs of milk"
Let everyone know I'll be late.
I can't do that, Dave.
task failed successfully
You mean "chat, how long is the line at the soup kitchen and is the bread line out of bread?“
"Comrade, dwell not on the selfishness of one's own hunger, but on the hunger of thy bothers. Remain hungry that they may eat."
"what's my purpose?"
"you open the app"
“Oh my god…”
Brilliant
@@ttred7621welcome to the club buddy😂
What the general public doesn't realize is how much of a breakthrough it was to make an MLP actually produce a text sequence that took context into consideration and produced a response consistent with it. An nobody though it was possible because OpenAI literally had to have access to entire datacenters worth of data an training to do it. From a technical point of view, LLM's are some of the most exciting advances into pattern matching and time function predictors that we had in AI for maybe 3 decades.
But this means jack squat in the real world. There is so much more complexity and context that humans and even animals need to process in order to make meaningful actions over the world that AI isn't even close to matching it. AI still needs to solve countless problems in order to even begin to approximate human reasoning: associative memory, higher function abstractions, body-sensory reactivity, self-regulated pain-reward cycles, and many more aspects of what makes a "human" response human.
We have trained a machine to repeat a bunch of sequences, and we are ascribing personality and intention to those repetitions in the same way we curse the gods when the roulette wheel draws black for the 10th time in a row.
We need to make much clearer what we want out of AI:
-If we just want a autocomplete on steroids then we need to stop talking about AI and expecting the AI to understand concepts or to know what "everyone" means.
-If we want the AI to resemble a human response then we need nothing less than to create artificial life, because without a body and self-intention the AI has no base ground for interpreting sensory data, so its not unexpected that it will simply spit back nothing more that what you put in. But giving AI a "body"... that is a whole other can of worms.
ugh the AI griftcycle is somehow even more tiresome than the crypto one. at least my employer wasn't trying to use crypto to replace me
I’m still convinced that the people who think AI is the second coming of Christ don’t actually use AI for anything remotely useful or important.
Yes I'm sure you're right and all the tech giants that are collectively investing trillions into AI infrastructure and adoption have no clue what they're doing
We aimed for centaurs: human brain + horse power for execution.
We got the inverse-centaur: horse brain taking decisions + humans running after it to fix its mistakes...
The normies think everything computers do is AI
openai, who have produced the most groundbreaking models every time, have described GPT4 as an "orca" and GPT5 as a "blue whale". so when GPT5 comes we'll hopefully get one last experience of that feeling of wonder, seeing it produce some unbelievable behavior, before we get used to it and realize how predictable and robotic it is.
@@tacokoneko I wish to agree with you. But unless the model/training changes fundamentally we still going to have the same errors, they will just be harder to spot because they will sound more convincing. They can improve the LLM's vocabulary 1000x, but no matter how fancy, LLM will still speak nonsense. In the end, it is still linear algebra and probabilistic distribution, even with all the scripts, Indians, and other things that they put on top to make us believe that the chat is actually reasoning.
"Maybe the real LAM is the friends we made along the way"
WHO THE FUCK WOULD LET AN AI ORDER THEM FOOD???
I will never not want to select what I want from a menu with photos and confirm everything is correct before ordering.
It's not like it takes that long to do it from your phone...
Imagine when an update creates a defect and the AI orders something that the user is allergic too...
Update: Version 4.1 fixes small error where we accidentally unlived a few users with nutrition sensitivities.
I run selenium RPA scripts professionally on external applications via selenium grid and, let me tell you, if you don't yourself own the external application and have control over its development, keeping these runtimes in sync is a stateful, unpredictable, hard-to-maintain mess. The UI becomes the system of record, but it's in no way ACID compliant and rollbacks are more or less impossible. Always call an API if you can.
Yeah because RPA is grift. Get a real developer job, RPA will be dead in a few years anyway
The OP is spot on.
RPA is more fun than APIs by 10x
I wrote my own little one that sucks but it's mine and it's perfect :)
@@hopelessdecoy Like an attractive woman on a first date: the sexier, the flakier
Hey do you know if the software(s) inside this thing could run on a normal Android phone? Or if it requires some kind of special chip that isn't on Android phones? Because I have a feeling they could've just write an Android application that would run the same exact code... The fact it is its own device is a good pretext for the price tag while almost nobody would pay even half the price of this device for the same thing but as an Android application, even if it was executing the same exact code.
"How else do you navigate a website"
Selenium using a chrome webdriver with python and beautiful soup library.
beautifulsoup has to be the best named library ever 😂 thank you for bringing it to my attention
@@nineephe You're welcome 😁
God bless
“I assume it’s playwrite” this man is psychic
Prewatched!
Just know things, think of the least resistance path to hack that crap together. Here you go.
wifugen watches coffee...may be told prime
Playwright*
It's pretty obvious actually when you know the state of the market and technology
The big linear algebra industry!
We aimed for centaurs: human brain + horse power for execution.
We got the inverse-centaur: horse brain taking decisions + humans running after it to fix its mistakes..
Tailwind catching strays 💀
Couldn't you just strip all styles before feeding it to the gippity
@@Oglokoog if i understand how playwright works the web documents are not fed into the AI. Playwright javascript is rather injected into the website to query document elements and trigger virtual clicks.
The main problem: hard coded scripts. When they first announced this i thought if this was true LAM it would involve some kind of image recognision and kernel level human input simulation that will work on any app or interface in foreground.
The LAM is the friends we make along the way
It hallucinated the response for how it knows where you are.
"Accessibility for AI" -- we have that. It's called an "API".
before: prove you're a human by solving this captcha
after: AI-accessibility
I would like to take a small amount of credit for “why is Eva everywhere”. I don’t know what I’ve unleashed on the industry but I’m proud as hell of her
Memelord AI job interview title is scary accurate
What would I get if I asked the AI to restock my fridge? A lot of bananums?
"I've replaced your beer shopping list with an assortment of vegetables, and here's a walking route for you to go get them, fatass"
underrated comment
Tomatums maybe
At this point, I've seen so many scams and useless products that are downright scams, I've developed an intuition for spotting it.
The issue is that I don't feel particularly offended and just ignore it and move on. But perhaps I should try to stop them and look into the things to warn others. Maybe that is my personal scam: The scam we made along the way.
I just sprinkled AI everywhere to get normal VC funding for my company which has a legitimate product (without AI) and it worked. VCs are like AI, they just do pattern matching, they don't really understand shit.
@@monad_tcpthat's both hilarious and sad at the same time
19:21 I couldn't brush of the uncanny backroom vibes to see the Amazon box in this empty room
12:30 sounds like 90's again. "Oooh, in the internet you say? Take my money!"
Prime is modest.
If he has a Masters in AI, he knows more about AI than Nvidias CEO, who is nothing but an Electircal Engineer and mainly business man who hires people to study AI.
Most people who study AI, literally just study LLM's and programing hoping to accidentally create an AI. It's not as cool as people think. Like Prime said, it's not magic.
The Rabbit shit is a scam and I hope that someone finds a way to take all 20million they made away from them. Nothing worse than manipulator preying on the people.
It's so damn frustrating for those of us in the general field, too. Nothing against LLM and ML people, but that's only a small fraction of the whole field, and places where there is really cool cutting edge ideas going on are just being ignored because they're not LLMs.
I dont think its right to discredit Jensen Huang because he seems to lack academic credentials. There are literal Turing Award winners with only a Bachelor’s degree.
@@ianzen youre right.
but im not discrediting him. i was using his creds to credit Prime.
"Valid certificate in AI" is going to become a scam, as well.
It was one Christmas and I bought EA game Jedi knight something. It took me 2 hours and my sanity to finally pass their 21 step captcha. It was insane.
Are you a robot?
This is perfect crossover, you two are delightful to watch
"Website design for AI accessibility" = APIs
App navigation could be fully automated if there was something like OpenAPI contract specification for UI stuff. Also OpenAPI(swagger) could actually make it possible to interact with that many apps.
All future rabbits:
User: order me breakfast.
Rabbit: Our Doordash rabbit is down for maintenance, instead we're connecting you with the Cloudflare Trust and Safety team.
circa 27:00 "To start off, I would love to know more about your background in tech.." That butt-hurt intro right there, negates any credibility of the remainder of their response.
51:49 Rabbit's LAM should be renamed to 'Lame Asinine Manpower' bc someone will be woken up to deal with the CAPTCHAs behind the scenes.🤣
Sending the html and letting a modern LLM figure it out is absolutely hopeless for the firseeable future.
That would require some understanding, and searching the internet for some website with a similar collection of words wouldn't work.
There are interfaces for machines to communicate with each other. We've been building them for like 60+ years? Maybe more? They are called APIs...
Imagine an honest man approaching these VCs and getting laughed out of the room because "if you ain't cheating, you ain't trying"
People on AI keep "peter molyneux"-ing themselves
Peter Molyneux was (big emphasis on was) a great game designer for his time and very influential in some genres even up to today. You could even put some of his bullshit down to getting overexcited and thinking interviews were brainstorming sessions. Most "AI people" making noise aren't even involved in training models beyond maybe a LoRA or two and merges.
I have a lot of problems with the baked beans salesman who lucked and half conned his way into software but for once I think the comparison is unfair to him.
@@MechanicaMenace Yeah Todd Howard is a better comparison
You could train an LLM against both html and a page screenshot to generate those playwright/selenium code. Inject changes into dataset, see how it adapts. I think it's very much possible and actually gonna happen at some point in near future, for something like that the tech is not the limit, maybe the processing power to train the thing is.
Jesse Lyu's background is in finance and marketing, yet says only those with some sort of valid AI credentials could possibly understand his gadget.
AGI is just 5 months away.
5months... Will age like spoilt milk.
What's most strange about this is that while integration is hard, some services are obviously get more used than others. This means implementing Uber, Amazon, and other big/huge names directly (API) produces more stable and faster replies.
They didn't even do that.
Ok I'm glad Prime has trouble with those dice captchas too, I thought I was going crazy doing those
These dudes saying their LAM is real reminds me of Nikola Motors claiming their hydrogen tractor was real truck, and not just something they pushed down a hill to make it look like it drives.
All AI roads lead to India
I want my fridge full of bananums
Good advert for the playwright part. That codegen sounds neat 😂
Your TH-cam channel is the best thing I have seen on the internet in so so long. Thank you for saying what my brain says every day at work. Ai (the bullshit bot variety aka llm) is a terrible solution to all of the problems in this video. 15 years software engineer. You complete me.
6:42 Well, as a Selenium enthusiast myself, I would actually love to have AI-accessible pages. Makes web-scraping orders of magnitude reliable. More bots to the god of bots!
AI in Germany can determine from a face picture, which party the person is going to vote for in the next election.
Can it tell if you're a Jew tho
@@rusi6219 Everybody knows you have to learn Craniometry for this task.
@@TheBadFred lifehack: drop pennies on the floor
New Foundational Model=Paying Humans in Fiverrr to act like LLMs
This is literally like something I created for a contractor to handle mass call volumes - use several nlp models to detect an intent from a given context, trigger an AWS lambda function to call a script and do a predetermined thing. It was utilitarian but NOT in any way "intelligent".
Fridges are a peculiar tools.
The smaller it is, the more things you'll manage to fit in them, and the bigger it is, the more you feel inclined to leave empty space, and pretend it's full when it's only half full.
Instead of paying for an AI Bot, why not hire a real person, a Virtual Assistant that is more than capable of helping you.
There tons of people from 3rd Worlds willing to do that job since US Dollar is superior compared to their money.
I know coz I'm one of them :3
The random locations when replying to "near me" could just be it repeating responses from the traning data. You search on google restaurants near me, google correctly replies with restaurants near you in texas, rabbit learns to say texas in reply to "near me". AI is great.
LAM is the friends we made along the way
omg the tailwind rant at 19:00 got me rolllling (love tailwind btw)
30:10 and people call Uber when they are too drunk to drive. Imagine trying to do the f dice counting while drunk or high.
I love writing passive agressive Jira tickets like: "stop breaking this software license." in some code that has been there for a decade.
I worked with some Asia based techpreneurs and "xxx is a marketing term" is the mantra over here. It's very typical to make a fully automated micro showroom as the shop front and then fully manual production room with hundreds of underpaid employees which would never ever be shown to the customer. It's called good business.
Last comment, Prime i found you this year. Been coding since 1994 when i was 9, and you brought back enjoyment and passion to it this year for me. Appreciate your vids.
The real LAM is the friends we made along the way
That’s literally the Octupus recipe app from the Silicon Valley 😂
25:51 - "We welcome you to invite experts with valid certificates from the field to further look into this" :: "and we would very much appreciate it, because we don't have any of them ourselves, all we have is a fleet of Playwright automation engineers."
38:50 I think the justification is that Google and Apple probably wouldn't sit by when they find a scam app suddenly blow up
The fact that we configure language models by setting out a list of rules in natural english and telling them to follow it is simply astonishing. These things are crazy smart.
Just because there are bad actors in the AI hype, does not mean AI hype isnt real. WHICH IS SCARY!!!
I hope AI ends up like NFTs.
Last year I thought machine learning was what I wanted to do, then I learned ACTUAL machine learning engineers were getting shunned for the generative AI prompt engineering crowd and all the promises that came with this new wave .
Actual machine learning is pretty boring , you spend months on a single project that may not see the light of day . The cool stuff is developing whole integration of AI through APIs
Now I’m probably going into cloud computing because it utilizes like all of my CS degree and it’s not hyped to the moon
@@ci6516 I bet all of those AI enthusiasts would fall asleep the second they get lectured on the core that makes AI even a thing. It's pretty much like with every new trending technology: People don't even know the basics of it, and yet make conclusions. Not to say I'm the guy who knows about it all, but at least I'm not hiding the fact that I have no idea what I'm talking about.
It won’t though but it’s not gonna be AGI hypetrain either
It's good that malcolm now pursues all this scammers.
My first experience with using LLMs was GPT-2 in 2019
people experimented with neural networks since 80. Of course it is not LLM in complexity, but kinda the same.
I'm pretty sure in the future SEO will not be a thing. Instead we're going to have ASEYO (Artificial Search Engine Yield Optimization) and it's going to be named ASEYO because it sounds like a korean hello and korean dev will create it.
18:55 the way Prime stares daggers at (I'm assuming) Adam Wathan (the guy that created Tailwind) had me rolling.
The other video Coffeezilla did on this topic shows how the Rabbit people were working on an NFT product until the market crashed, and then they quickly pivoted to Rabbit.
Please take a look at unreal's new "inclusive coding standards". We now should be naming things as they, them, and their!!!
Can't use execute, abort, native, blacklist, whitelist because, you know... its offensive??
The fastest bender of all time, the one that stocks your refrigerator fast fast
33:16 I wonder if they don't have a GPS on this thing because it would take too much power. They use the Internet for localisation. For instance on a vpn from home, the work laptop thinks it's on a different city.
Approximate Location: Okay, I'm pretty sure you are on earth
AI = An Indian🤣🤣
the LAM is the friends we made along the way
18:16 😂 I laughed so hard when u blamed Tailwind for class overload hell in HTML, that can't fit in AI token. Yes, tailwind classes are useful, but it's a pain trying to create varied animated ui elements. It's probably just easier to create 1 scss mixin function with multiple properties, that can be reused. I have a love/hate relationship with Tailwind atm 😂
I arrived to hear information about Rabbit AI. But instead I found out about people trying to start a "Cult of the LAM".
Imagine a future where an ad needs to be read by the device first before any useful actions 🤖
Bro I just got a click up ad on this video and the WHOLE ENTIRE APP is "things that suck about jira" wtf is this lol. Not a single mention a clickUp feature just "jira sucks and we do it better" HOW DO YOU DO IT BETTER?
Like, I hate Jira but you told me nothing about YOUR product
Jesse going for the Sam Bankman-Fried "useless fat friend on the couch" aesthetic hoping it adds a zero to the raise.
They'll be using something like langchain agents to trigger commands and openai apis for the "chat" part of the system.
The funny thing is they are so stupid in a real world application and as a developer myself if i were to make this concept it would have been so easy using chatgpt but they messed up going the playwright route instead of working with companies like doordash and implementing a Application interface allowing the Language model to turn words into actions.
In the end it turned out the real LAM was the friends we made along the way✨
This could've been a fantastic accessibility to for the blind
Multics actually was created and Wikipedia says Multics 12.8 was released August 10, 2023. UNIX came to be because they felt Multics was taking too long and the hardware requirements were too high. (i.e. They felt it was over-engineered)
the random chatter at 40:43 saying :"press gg^vGd" made me laugh out loud
As long as business can't properly explain the business needs we are save
rabbit choose to be a scam, that’s their call. but don’t call llm powered app chatgpt-wrapper. it’s like calling backend database-wrapper and web app html-in-a-specific-order
That's exactly why my initial thought when I saw this device was "AI navigating through UI, hell no. What a waste of resources. Create an API platform...". UI navigation is such a slow and unreliable process...
I twould have been hilarious if he said, "That was delicious, check the fridge..." and then it responded with "No."
I remember that their ads were everywhere on Facebook. Funnily enough, no one knows what the product was or what it was supposed to do. They never answer those questions.
The problem is that less than 1% of people who are currently trying to be a programmer actually loves it and want to be a master of it, so the junior or even mid-level market is still going to be terrible for the foreseeable future.
Is rabbit subscription based? Because using the api costs $