Posts on X before Feb 4, 2024, 2:36:36 PM UTC

and doomer's justification for this is "because this is better than the potential of p(doom)" it's literally a psuedo-religious threat of 'the end of the world', the only cure being?... more rationing, inflation, bureaucratic friction and widespread pain. bingo.
Despite what the doomers say deceleration doesn't equate to "flowers puppies and everything nice", it's more rationing, inflation, bureaucratic friction and widespread pain.
"within this decade" sounds like a long time, but reality is he means in less than 6 years + factor in he's being pretty safe with his estimates so it will probably be 3-4 years if things stay on track with growth % increase
NVIDIA CEO Jensen Huang says computers will pass any test a human can by the end of this decade
i sincerely think apple made the 1st gen apple vision pro bigger than it needs to be because: - huge selling point for 2nd gen - they know only the fanatics will buy 1st gen - they'll do what they do best; "we know you loved apple vision pro, and we've heard your concerns..."
I mean all you need is to make a script that has mouse/keyboard functionality and simple "complete this puzzle" logic. allow the LLM to diagnose the question, move the mouse/keyboard if need be... and... solved. we'll see this in mass action in probably a month or two from now
Qwen-VL-Plus is SACARY good! (better than GPT-4V) here it is casually solving Recaptcha! - You don't have to give any additional instructions other than 'Solve it.' - It can even mark the exact position of the objects it is looking for. ↓ it can do so much more
Media attached to this post
👀 no longer do you need to send inspo images to an actual artist and hope they can draw in that exact style. there go a bunch more jobs/freelancers
The Glif StyleHunter browser extension is now LIVE - link below find cool image styles on the web --> right-click-remix them with your prompts --> quickly get a new image in the same style it's a super fun way to get inspired & generate with StableDiffusion *on the go*:
chances of openai becoming based and going full mask-off uncensored with a "warning" for generations instead of blocking requests? would sama even be allowed to do this?
The reason I made this image is because openai is actually doomed i know for a fact that they are not going to make it someone will take all their users because someone will actually serve them its a matter of time your moat isn't built yet back to lobbying before its too late
It's so crazy to look at content of any kind from 2+ years ago and realize... there was no mainstream AI tool to easily produce this There was no ChatGPT, LLMs (of today's quality), MidJourney, etc
EU seems like the worst of both worlds. USA's people can do "whatever they want" (to quite a degree), and so this brings innovation otherwise impossible through heavy regulations. Same for China (replace people with gov). EU can't effectively do either, so they play catch-up.
Germany, France and Poland announce the ‘Weimar triangle’ for artificial intelligence
Media attached to this post
to put this in perspective, it's like asking a Russian grandmaster in chess what each of the pieces are called in Chinese "because a good chess player should know it in every language". how about we just do the hire based on merit instead of the ability to make sorted lists?
Today I learned that a Kaggle competitions grandmaster was unable to get even to the screening interview stage for a low-level AI position at Google, the parent company of Kaggle. I am sure they will select a candidate who can perfectly sort a linked list on a whiteboard. And tha… MoreLess
Today I learned that a Kaggle competitions grandmaster was unable to get even to the screening interview stage for a low-level AI position at Google, the parent company of Kaggle. I am sure they will select a candidate who can perfectly sort a linked list on a whiteboard. And that, in a nutshell, is why I think that the job screening process in tech is absolutely f***ed up.
Truly ahead of their time. They knew about the Llama models before anyone else. Let's listen to these prophets and truly make use of Llama 3 when it comes out 🤝
Media attached to this post
for future ref, the recommended reading list at the back of the manual of The Sims x.com/FoolOfD/status…
Media attached to this post
I truly believe that one of the reasons GPT-4 is so bad at following directions (specifically to coding) is because of all the safety precautions it has in tact. ...if you train it to -not listen to certain asks-, guess what happens? This affects other areas of task completion.
openai is going to ultimately fail they lost as soon as they produced software that refuses to do what the user asks it to do. that sort of trust won't be regained
unfortunately it's going to get a lot easier, higher quality, faster, less expensive to generate, and a waaaay more mainstream. this isn't even 1%. absolutely nothing we can do about it other than prepare ourselves and our loved ones mentally for what this will bring to fruition
The true answer to this question (that to my surprise VERY few seem to realize) is in this Taylor Swift song. As the dad of two amazing girls I hope to live to see this effing BS change once and for all. music.apple.com/us/album/the-m… x.com/AIandDesign/st…
Until a Vision Pro headset is the price of an iPhone, I don't see mass adoption of it. Maybe in 2+ iterations of it, we'll see both an increase in quality/utility AND a decrease in price. Then again, if you compare it to a Macbook instead of an iPhone...
there are people stuck on a cruise for 9 months, documenting everything on tiktok. imagine the AI breakthroughs that will happen between their boarding and final destination... if there's ever been a time to build, it's now.
I love talking about health in the same way everyone used to talk about crypto/NFTs "we're so early" honestly gives me latest AI trends vibes
Ice pack on your balls in the sauna This is how you spot a Huberman type in the wild There is no one else doing this We are so early
Insane that in all those 2200 years, 1 of them didn't tip over and cause a domino effect.
It is believed that the Terracotta Army, referring to the soldiers guarding the tomb of Qin Shi Huang, looked like this when they were first crafted around 2200 years ago.
Media attached to this post
This seems like a recurring theme with anything that Elon touches. You think you've seen the height of an industry, and then he storms in and shows you what ACTUAL progress looks like. Every. Single. Industry. He. Touches.
I don’t understand. I just don’t. Robotics has been a thing for decades. Decades of research, and Boston Dynamics made a kinda interesting dog? Why does nothing useful happen until Elon decides to do it? What are all the researchers up to? x.com/elonmusk/statu…
The worst part is realizing that after this documentary was published, (most likely) the guy just watching Google was punished along with his next 2 generations
This is an eerie computer lab in North Korea
OpenAI ChatGPT really needs fuzzy searching through chats, both active and archived. That would make me use it 1000% more often.
Going to be diving into @openchatdev's new OpenChat 3.5 1106 GGUF through LM Studio today and trying it out with an Obsidian Vault using Text Generator. Should be fun. Stay tuned.
Take care of your data now.... ...and it will one day take good care of you.
The year 2024 in AI will be about data quality. Data quality changes everything. Here's what the author of Mamba, author of a famous architecture rival to the Transformer, said: The architecture stuff is fun, making hardware efficient is fun. But I think ultimately it's about d… MoreLess
The year 2024 in AI will be about data quality. Data quality changes everything. Here's what the author of Mamba, author of a famous architecture rival to the Transformer, said: The architecture stuff is fun, making hardware efficient is fun. But I think ultimately it's about data. If you look at the scaling law curve, different architectures would generally have the same slope, but they're just different offset. Seems like the only thing that changes the slope is the data quality.
The internet seems to be accelerating on X ever-increasingly... When's the last time you saw cutting-edge tech discussed FIRST on any other platform? It always starts on X.
I'm blown away by Mixtral 8x7B, the new sparse mixture-of-experts model from Mistral AI. It outperforms Llama 2 70B and GPT3.5 on most benchmarks... and it's open source.
@NickADobos I remember reading that you said Custom Instructions are taken into consideration with custom GPTs, but I've been testing extensively with your Grimoire / my own throwaway GPTs, and they definitely don't... am I missing something?
Even a 2% decrease of optimal hydration can lead to a significant debuff to mental performance. Stay hydrated.
AI isn't just a tool, it's a new colleague. Communicate. Collaborate. Innovate. Together, you're unstoppable.
I've been using Bing instead of Google as my default search engine for over 6 months now. Tried Google again just now. Yeah, definitely not going back. Bing is superior now. I never thought I'd be saying that. EVER.