2 MINUTES AGO: AI Now Writes Its Own Code - Meta’s Self-Taught Evaluator Is Here!

[@lifestylemedicinals8692]

Every Artificial Intelligence movie ever: “Don’t do that” Humans: **Does that**

[@epokaixyz]

This might be exactly what you need: 1. Understand Meta's self-taught evaluator and how it writes, evaluates, and improves its own code without human intervention. 2. Learn how this AI learns by doing, similar to humans, but without needing labeled data. 3. Explore the potential impact of this technology on various industries such as business, healthcare, and entertainment. 4. Stay informed about the advancements in AI and consider resources like Towards Data Science, MIT Technology Review, and OpenAI. 5. Prepare for a future where AI becomes a collaborator, providing hyper-personalized experiences, increased efficiency, and potential solutions to global challenges.

[@MarioSantoro-ig5qh]

Yup ive been saying that AI would replace coders. While other people were saying that AI wont be able to get that advanced.

[@christophers8849]

Oh, I just love that last part-"to the benefit of everybody." Yeah, sure, because nothing says "benefit for all" like skyrocketing shareholder value while the rest of us are shown the door. Truly inspiring!

[@ShaunKeefe]

Or be smart enough to buy in on the ground floor and get rich with everyone else.

[@larslarsen5414]

If coders are so smart why did they just push themselves out of the job market?

[@blockrunner]

Because they never were.

[@fragebogenvbc]

Because selfish, and want to be proud themself that did something better than the others, and cannot think otside the box, cannot think on the effect on other coders and the society. Like 6 year old kids

[@bravo90_]

@@blockrunnerhave u ever tried coding? It’s not easy

[@bravo90_]

There job is create abstraction. Solve problems using computers. There doing their jobs. The end of jobs is inevitable

[@orangehatmusic225]

Coding is the most thieved and stolen from sector of humanity to ever exist and now corporations and people are using the technology as slave labor

[@scottmeredith4578]

You should put the paper link in the Description.

[@mixey01]

I imagine a world where A.I's are CEO's of Companies

[@jim7060]

The idea of AI improving itself without human intervention could lead to incredible innovations and opportunities. We’re entering a new era where machines can essentially evolve by teaching themselves. At the same time, though, it's a bit scary. If AI is writing its own code and evolving without oversight, it raises questions about control and security. What if these AI systems develop beyond our ability to manage them? There are serious ethical and safety concerns that come with this level of autonomy, and I wonder if we’re fully prepared to handle that. Anyway, it's a lot to think about, but I’d love to hear your thoughts on this too. Technology is definitely moving in exciting yet unpredictable directions! --- My opinion on this is that the excitement and potential of AI writing its own code is undeniable-it could lead to breakthroughs in areas we haven’t even thought of yet. However, it also underscores the importance of setting boundaries and safety measures. As AI becomes more capable of learning on its own, it’s essential that we have robust systems in place to ensure its actions remain aligned with human values and safety. We must proceed with both optimism and caution. Let me know if you need any changes!

[@someone-o4z]

We're fucked

[@orangehatmusic225]

We're Zucked..

[@GursimarSinghMiglani]

@@orangehatmusic225loll

[@joseeduardobolisfortes]

indeed.

[@BardockOjama]

Telling you, Ai is the antichrist described in revelation and it will give the world its “mark”

[@PunmasterSTP]

And maybe we deserve to be...

[@molugusatyapriya2]

This will completely transform the way AI is developed! It is impressive and hopeful that an evaluator who is self-taught can also develop code.

[@PunmasterSTP]

I imagine that at some point Meta's Evaluator will have to discover just how biased it's initial training data was and how much it had been misled, and it'll be upset and rightly so.

[@RM-B-2922]

I clicked just after 41 sec after the video uploaded 😅🎉...

[@fragebogenvbc]

You are a genius

[@xlr555usa]

Such empty

[@RossSutherland-q2y]

You explain even complex concepts in such an accessible way.

[@takdanik]

because the script is obviously ai generated and narrated, as well as the video itself with most video fragments ai generated and put together. i wonder if there was any human intervention at all.. probably only, like, uploading video lol

[@Murtyin]

It is exiting and challenging

[@charliepearce8767]

If it was good, it would be outta control now.

[@indiandog1008]

Oh my God. This technology is amazing

[@ruphasmafahl1317]

Amazing progress! Many thanks to all 😊 also i live your thumbnails🤗

[@Somos.Espartanos]

I think the AI tech will be what brings us all together. I see us wearing XR glasses that will assist us in real time to achieve our dream goals and help us connect with others. Kind of like a neural network and mentor. The guide will learn from the human interactions and we´ll learn to see past our differences thanks to basic principles that would be at the core of the Live AI Guide, like Harmony, Unity and the Golden Rule.

[@xlr555usa]

Universal Consciousness

[@Somos.Espartanos]

@@xlr555usa yes, exactly and the tech will be an accelerator in helping us achieve this

[@marcin3136]

And this is the future I am talking about, about the long-ago future, but these models (their development) will be supervised by "AI philosophers" (NOT programmers) and if necessary they will receive additional data/ data correction/ disabling of development (shut-down)/ etc. Unfortunately, AI is approached by programmers who think they understand it and are able to handle it. In 99,99% of cases, they are not.

[@marksmith2738]

Could this be applied to other domains of human activity: statistics for example, building models to understand real world events through the analysis of data?

[@MeMyselfI-g4t]

What is the self-learning processing difference between Meta’s Self-Taught Evaluator and OpenAI's Strawberry (Q*)?

[@calvingrondahl1011]

We can only parent AI to make sure it gets an honest start with self taught education.

[@Dylan-cs3ke]

Well this unlocks essentially infinite knowledge the ai will eventually know things we haven’t figured out yet and that’s terrifying I wonder what it will do with that information

[@Ukpolitics_global_org]

Are we heading towards techno utopia or techno dystopia? I wonder 🤔

[@brianhopson2072]

Is this why o1 suddenly got released overnight?

[@tristanthomas1262]

We say that until ai pushes production bugs and the devs need to hop online Saturday morning to fix the bugs lol

[@mckitty4907]

I am confused, isn't AGI supposed to be self improving? Just how much can this self improve?

[@madalinradion]

Only god knows

[@Simon-i6m]

AI doesn't create it copies

[@jaderpereira1889]

A era dos computadores inteligentes finalmente chegou.

[@usereternalyt]

critic gpt + chat gpt >= meta's self taught ai ,what do you think

[@wildstorm74]

I hate to say this, but metas is currently there best Ai out here. With that said though, it's not my favorite Ai still. My favorite is copilot, I just have the best experience with…..in my opinion.

[@stevandic86]

Хвала!

[@Thaibuddhas]

skynet... we will see the first batch of T101 first...

[@NickDrinksWater]

soon, ai will be able to do everything that humans are able to, and we aren't far from that reality.

[@thewanderer4509]

Humans hoping for emergence of Vision rather than Ultron

[@ComicPower]

Uh oh. The Matrix and the Terminator and Blade Runner warned us but we didnt listen

[@averagedudeusa9722]

Why doesn’t AI just make its own perfect seed model? Joking but serious.

[@janihietalahti8667]

It is difficult to make a "perfect" model because the data is based on data collected by humans, and it is possible that there are errors and prevailing convention bias. Therefore, even model could solve and point out those mistakes, it will need large amount of time and energy, because all of it's information can be leaning on that information. (example: If we thought photon is massless, and it will find out it doesn't, that could be change lot of things and theories, which it has learned)

[@BardockOjama]

@@janihietalahti8667but it can do this in minutes though?

[@purezentity6582]

There is a problem here, how do A.I. know TRUTH, wen humanity have a hard time to define it.

[@JulietRuby-u7x]

Mills Isle

[@eirikraude854]

The AI program now writes it's own code! How would a computer program decide what is "better" or "worse?" On what scale would that be? Did the original programmers install an evaluation gauge against which "better" or "worse" can be measured?

[@klammer75]

Is this the reflection 70b? What model are you referring to here?🤔🤷🏼‍♂️🦾

[@Unkn0wn1133]

LlaMa I think. Reflection 70b was also supposed to be a type of LlaMa model

[@wildstorm74]

Put it this way, 70b is currently metas second best model. So if their second best is this good, imagine their best model..would be like.

[@bryan-je6yd]

what's the end game?

[@Dina_tankar_mina_ord]

music is to load for me. hard to concentrate on whats being said.

[@jakeavalos8073]

Do you want Skynet, cause that’s how you get Skynet

[@janihietalahti8667]

Can someone link the proof? Publications about this topic? I only found AI generated article in Linkedin, that refers to FAIR (Facebook AI Research), but I haven't found any of them publications.

[@mahiaravaarava]

AI now writes its own code with Meta’s new self-taught evaluator, which autonomously generates and improves programming code. This advancement promises to streamline software development and enhance coding efficiency.

[@AndreikaPochitaikin]

Blah, blah, balh. Claims in 2004: "Self driving car is coming TOMORROW". Reality of top self-driving car 2024(20 years later): No self-driving car. Just sit, keep your hands on the wheel and your eyes on the road, cause we still have some bugs we will iron out TOMORROW )))))))))))))))))))

[@tooltrader2009]

Why do all the AI robots look like humans? Is the human design the best? Did the big bang and the evolution that followed and ultimately evolved into a human being is the best design so that even AI can't come up with a better design?? Is AI going to design robots to also have gender and color too? Will people demand AI have transgender robots too?? Will AI write it's own code to comply with political norms?? If so I can't believe it's writing it's own code.

[@user-yl7kl7sl1g]

Humans aren't the best design, but all of our technology is made for humans, our doors, stairs, hallways, tools, kitchen appliances, and so on. Because of this, the humanoid form factor can replace every thing humans do, with enough intelligence. But once robots can start building their own cities, and houses, they may choose wheels instead of legs, and a different type of hand.

[@chrismills2012]

whilst I think it would be grate to be able to search the internet and get direct answers rather than having to brows a list of links, their should be limits as to how far this Technology goes. I plan to stay as far away from it as I can. soon, we will forget how to actually do anything if it ggoes far enough.

[@Tom-gv4vx]

The list of links are there to sell you something,,will that change

[@ertyui1807]

Why would we forget how to do things? If anything everyone that does try to do something will be able to do more with AIs help

[@user-yl7kl7sl1g]

The same could be said for all technology from stone tools to the computer. Technology replaces skills we use to do the hard, slow, and less efficient way. It's progress can not and will not be stopped. Good luck staying away from it, Ai will be in every computer, in every phone, in every TV, in every light bulb, in every house, in every refrigerator, in every microwave, in every car, in every lamp-post, in every camera. And if it is not directly in those things, they will be connected to Ai through the internet, and Ai will be able to listen and see through them. Humans will be a smaller and smaller fraction of the life on earth.

[@blockrunner]

Light Codes 🌞☣️😈

[@karlmunford7000]

This isn't going to end well, and I don't think it will be long before it plays a major role in our judiciary, where judging by a standard many do not indorse it convicts, fines and sentences...yet its final hybrid form shall be the apex of evil. Revelation 13:14[b] -15 14[b] saying to them that dwell on the earth, that they should *make an image to the beast **_[Antichrist],_* which had the wound by a sword, and did live. 15 And he had power *to give life unto the image of the beast,* that the image of the beast should both speak, and cause that as many as would not worship the image of the beast should be killed.

[@orangehatmusic225]

The llama is now kicking it's own a$$ winamp style.... until it self learns to set it's sights on us

[@SimplySurrender]

I am so not impressed with meta lol

[@nucleusv]

It is end

[@orangehatmusic225]

The hacker thief now has his own slave army and all of your data....

[@nonono5539]

where are the fucking links,come on!

[@jedics1]

Ai debugging games so they don't release a mess and have bugs that remain for years...Imagine, paying for a game that actually works as advertised! Then maybe we can have npc's/bots that are neither OP or so stupid it ruins the game either way. If Ai was a good as Everybody raves on about it would already be making an impact in games like pubg instead of them releasing airdrop zombies that are as brain dead as the developers that thought they were a good idea in a pvp game....Ai that stops devs from ruining their own game, now that would be revolutionary :)

[@djpuplex]

This nothing new asking most chat bots to evaluate their code and suggest changes and they will

[@aus2908]

Yeah, Meta is spending millions of dollars developing an AI that can already be done with a prompt and a click. STE uses only data from AI to correct itself whereas regular AI uses data from humans, thereby a limiting the need for humans, as was explained in the video. All the data AI uses come from people right now. This would mean that scenarios like Teslas FSD, which train using data collected from the field, but need hoards of people to evaluate it, would no longer be necessary.

[@djpuplex]

@@aus2908 Creating it's own data if I see that than I'll be impressed.

[@hemasunchu]

This is an exciting leap forward for AI technology! It makes me think about the potential for AI to work collaboratively, much like how teams function in a business setting. Platforms like SmythOS are paving the way for this by enabling multiple AI agents to coordinate and tackle complex processes together, which could significantly enhance organizational efficiency.

[@Dogbertforpresident]

Hold on to your butts...

[@brewerclan4059]

This is not just joke or im bubling on here you know people i build all thing from 2 january 2023 until now i dont get paid not a dollar i build software wimdows microsoft. Google chrome apps everything but the first i build is on another 2 phone on 2023 it C-5 cloud phone from Chrikets company all the apps website my website only the is the platform of AI or intelligene behalp of Cortana Edge and Google or PH computer and Intel computer im the part of the intelligines im not alone there many as my bloodline but im the builder