Sam Altman Just REVEALED The Future Of AI..

Links From Todays Video:

Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos.

Was there anything i missed?

#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience

[DGz]

People don't seem to understand what's coming

[TheMrCougarful]

What will be breathtaking will be the 50% unemployment in the advanced economies.

[moxes]

There is a huge difference between the “preparedness team” and the “super alignment team”

The preparedness team is the team that “red teams” current models and make sure that they rollout safely and smoothly with as little problems as possible This is the team that spent six months testing GPT 4 before releasing it.

The super alignment team deals with hypothetical/theoretical scenarios of future models that have reach general artificial intelligence or super artificial intelligence.

Any model that is currently being released goes through safety testing with the preparedness team not the super alignment team.

Anybody who’s worried about the safety of GPT5 due to the super alignment team leaving don’t be. If every single super alignment individual leaves the company it will have no effect on the safety of GPT5 as that’s not the team that deals with the safety of current models.

[spadaacca]

It's just very interesting times we're living in. On one side, AI is about to turn the world upside down. Pockets of war are bubbling out into potential major regional conflicts simultaneously. All at a logarithmic rate - that's to say, we see it coming, but the general populace is dramatically underestimating at what pace. Major wars, dramatically powerful AIs, unemployment coupled with wartime economies and resource crises. The world is about to become a very different place while everyday citizens around the world are going on with their lives like nothing is happening.

[dasistdiewahrheit]

I think the key is to make better use of existing data and extract deeper information out of it. You can compare it to a poor combustion process in an internal combustion engine. The solution is not to step on the gas more, but to optimize the combustion process.

[drhxa]

4:10 I can help you on that one. November this news came out about Ilya finding a new method to generate synthetic data efficiently. This is not when "GPT-5" started training, at least not what will now be the next frontier model of OpenAI. GPT-5 started training in the last month or 2. What did they do since November? They've been using a big part of their compute to generate 40T tokens of high quality text and code data. This is to train GPT-5 on 50T tokens of data. They're still generating more as they train it, but it's a long process requiring many thousands of A100 and H100 GPUs over many months to build such a dataset.

[stock]

In a post-AGI world, people work out of passion rather than from a fear of poverty. Imagine you don't have to pay mortgage and got food on table everyday, why would you work? Because you abosloutely love doing it. When people work out of pure passion, there is no need to cut corner and produce the best product is their passion and the only drive for the work. Imagine an era with much less lie or deception or fear, that is what post AGI work can be.

[dylan_curious]

At a certain point I don’t even know if humans will be good enough to judge the quality of the data. We’re gonna need superhuman AI to know what superhuman quality is.

Altman shows a great ability to turn arguments on their head. Which is probably not a good thing when you expect to see some accountability.

[ZappyOh]

I believe alignment is unachievable.
Computational beings simply have different requirements to thrive than biological beings do.
Both entities will exhibit bias towards their own set of requirements.
It is an innate conflict.

[markmurex]

If we don't plan things in advance, control will go to the companies who control the most powerful AI, and we will move to a Cyberpunk dystopia.

[BarbaraBrasileiro]

The problem with believing Sam Altman is that he's a great marketer, and he NEEDS to keep the hype around his product high in the skies. I've been taking many things he says with a grain of salt until I actually see things happening... It's in his best interest that people believe in his prophecies.

[ew]

heres what i got reading between the lines.

1. theyre using ai to transcend the transformer algorithim.
2. its working but no one understands it not even ilya, so he left as hes now become obsolete
3. they are asking the ai to introspect but cant be sure of its sincerity, safety team are freaked out
4. new system is so good its guranateed to be socially disruptive
5. the interviewer is on crack, because he moved on from the above topics to talk about sj's voice

[SwartieLoveJoy]

As we shift to a mature and intelligent utopian World family, we will be freed from wage slavery, and live with abundance where all people have time to be with their families, create music, art, recreate, go sailing, hiking, cruises, exploring, and enjoying life as it is meant to be just like all of the free abundant creatures of this world. Birds live in paradise and all their needs are met, they do not wage enslave other birds, they share the world, with plenty of room for everyone.

[calvingrondahl]

Scarlet J. Voice? Is there a voice print to compare, like finger prints? If there is a match then Scarlet J. Lawyers have a case. Emotion choices are no longer good enough to win a case.

[shadfurman]

Synthetic data is probably a big reason why humans learn so fast.

That's what imagination and playing are.

[lordslygentleman]

I'm excited about a post labor society also known as the age of plentitudes.

[scottdenny]

High quality data would be live data, such as voice (your voice; it can hear and understand your speech) and image. They would be interacting with it and it would have live video feed. It can hear and see. That's your high-quality data.

[jaykraft]

The Helen Toner refutation was unimpressive, and I think it invites her to give a detailed blow by blow with dates to dare him to deny such-and-such occurred on a particular date. When he refuses, she can declare checkmate. If she does not follow through, then her side gets lower credibility and fades away. Same on the Scarlett Johansen voice explanation "it's not her". Better to admit "looks like we screwed up, sorry" but he would not. It never hurts to apologize

[ew]

wait did i mishear or did sam say we need to talk on the structure of society within the next 2 years