Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

GPT-5.6 Sol has been used to solve open problems in mathematics. So why was it struggling with ARC-AGI-3, a benchmark of 2D puzzle games? We investigated. The harness was not letting it remember what it had learned. We found that enabling two API settings tripled our scores with 6x...

OpenAI's profile picture

OpenAI

5,077,561 subscribers

1,128,096 Aufrufe • vor 2 Tagen •via X (Twitter)

Wissenschaft & TechnologieBildungNachrichten & Politik
LIVE

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

Sol, Terra, and Luna, our GPT‑5.6 family of models, are starting to roll out now in ChatGPT, Codex, and the API.
3:26

Sol, Terra, and Luna, our GPT‑5.6 family of models, are starting to roll out now in ChatGPT, Codex, and the API.

OpenAI

4,460,976 Aufrufe • vor 22 Tagen

Introducing autoresearch with GPT 5.6 We had GPT 5.6 Sol reproduce the key findings from “Towards Mechanistically Understanding Why Memorized Knowledge Fails to Generalize in LLM Finetuning” Compared to GPT-5.5 and even Fable 5, GPT-5.6 stayed more focused on a few, critical experiments and spent less time on peripheral details. It also asked fewer “clarification” questions and independently resolved ambiguities instead of pushing them back to us OpenAI pushing the boundaries of the automated research loop with models that don’t have handcuffs
0:23

Introducing autoresearch with GPT 5.6 We had GPT 5.6 Sol reproduce the key findings from “Towards Mechanistically Understanding Why Memorized Knowledge Fails to Generalize in LLM Finetuning” Compared to GPT-5.5 and even Fable 5, GPT-5.6 stayed more focused on a few, critical experiments and spent less time on peripheral details. It also asked fewer “clarification” questions and independently resolved ambiguities instead of pushing them back to us OpenAI pushing the boundaries of the automated research loop with models that don’t have handcuffs

alphaXiv

85,137 Aufrufe • vor 18 Tagen

McAvoy on the team locking things down defensively since their 6-game losing streak: "It has to be our strong suit, because that's the kind of team that we have. So, I think that was why it was probably the most frustrating when we were going through that pocket, was we were defending poorly and that has to be our strength. We've done a much better job the last couple games with that."
0:38

McAvoy on the team locking things down defensively since their 6-game losing streak: "It has to be our strong suit, because that's the kind of team that we have. So, I think that was why it was probably the most frustrating when we were going through that pocket, was we were defending poorly and that has to be our strength. We've done a much better job the last couple games with that."

Bridgette Proulx

11,090 Aufrufe • vor 9 Monaten

"What is P vs NP and how does it relate to human knowledge?" Avi Wigderson (Turing Award & Abel Prize winner): "In a very basic way, all the problems we want to solve are really NP problems. A subset of it is the problems we can solve. So the question of P versus NP is whether we can solve all the problems we want to solve, whether we can know everything we want to know. This is basically about the limits of our knowledge. You can think there is no limit to what we may want to know but I claim that essentially in any human endeavor, what people are searching for have the property that when a solution is found, it is easily recognized. NP are all problems that we can honestly say we really want to solve and P are those that we can solve. Whether they are equal is just the very fact that a solution can be easily recognized would imply that it can be efficiently found. That means that we can know everything we ever want to know, it's a fundamental question about human knowledge."
1:00

"What is P vs NP and how does it relate to human knowledge?" Avi Wigderson (Turing Award & Abel Prize winner): "In a very basic way, all the problems we want to solve are really NP problems. A subset of it is the problems we can solve. So the question of P versus NP is whether we can solve all the problems we want to solve, whether we can know everything we want to know. This is basically about the limits of our knowledge. You can think there is no limit to what we may want to know but I claim that essentially in any human endeavor, what people are searching for have the property that when a solution is found, it is easily recognized. NP are all problems that we can honestly say we really want to solve and P are those that we can solve. Whether they are equal is just the very fact that a solution can be easily recognized would imply that it can be efficiently found. That means that we can know everything we ever want to know, it's a fundamental question about human knowledge."

Ryan Peterman

20,069 Aufrufe • vor 1 Monat

🚨 PRESIDENT TRUMP JUST NOW: "The main thing is that Iran will not have a nuclear weapon!" "They fully agreed to that with strong POLICING powers and they won't have a nuclear weapon, which is what it was all about because they probably would have used it if they had it." "So we had two big moments when we terminated the JCPOA, that was the Obama deal, the Barack Hussein Obama deal." "And when I terminated, that was very important because it was a road to a nuclear weapon. It was a horrible deal for the United States. It was a deal where billions of dollars was given to Iran."
0:53

🚨 PRESIDENT TRUMP JUST NOW: "The main thing is that Iran will not have a nuclear weapon!" "They fully agreed to that with strong POLICING powers and they won't have a nuclear weapon, which is what it was all about because they probably would have used it if they had it." "So we had two big moments when we terminated the JCPOA, that was the Obama deal, the Barack Hussein Obama deal." "And when I terminated, that was very important because it was a road to a nuclear weapon. It was a horrible deal for the United States. It was a deal where billions of dollars was given to Iran."

Eric Daugherty

76,585 Aufrufe • vor 1 Monat

lando: “it obviously was not a bad decision for lap 1, it worked out in a way. at the same time it was the wrong decision. so we have to accept that.. but i also pushed for it myself quite a bit. so we take it on the chin” “but we had some problems and we had a failure in the end which put us out. so yeah, put us out of misery in a way” 🙃🙃
0:46

lando: “it obviously was not a bad decision for lap 1, it worked out in a way. at the same time it was the wrong decision. so we have to accept that.. but i also pushed for it myself quite a bit. so we take it on the chin” “but we had some problems and we had a failure in the end which put us out. so yeah, put us out of misery in a way” 🙃🙃

ray

209,687 Aufrufe • vor 2 Monaten

Earlier this month, an Erdős problem that had been open for 60 years was solved with help from GPT-5.4 Pro. What happens now that AI is getting good at math? OpenAI researchers Sebastien Bubeck and Ernest Ryu join host Andrew Mayne to explain what changed and what it could mean for the future of research.
43:28

Earlier this month, an Erdős problem that had been open for 60 years was solved with help from GPT-5.4 Pro. What happens now that AI is getting good at math? OpenAI researchers Sebastien Bubeck and Ernest Ryu join host Andrew Mayne to explain what changed and what it could mean for the future of research.

OpenAI

472,511 Aufrufe • vor 3 Monaten

Last weekend I checked in with Stipe Miocic , but it would have been unfair to not revisit our trilogy. We did that and I’m releasing the video in :10 mins. What a guy, what a fighter check it out at 6eastern/3 pacific. It was fun
0:54

Last weekend I checked in with Stipe Miocic , but it would have been unfair to not revisit our trilogy. We did that and I’m releasing the video in :10 mins. What a guy, what a fighter check it out at 6eastern/3 pacific. It was fun

Daniel Cormier

174,814 Aufrufe • vor 3 Jahren

We found what was left of our last bike. 💔 It was bombed—along with the place we stored it. In Gaza, movement itself is treated like a threat. From donkey carts to cars to bikes, anything that helps us move becomes a target.
0:32

We found what was left of our last bike. 💔 It was bombed—along with the place we stored it. In Gaza, movement itself is treated like a threat. From donkey carts to cars to bikes, anything that helps us move becomes a target.

gazasunbirds

10,677 Aufrufe • vor 1 Jahr

One of the reasons why I left the USA was because of woke culture. I didn’t understand why I had to wear a name tag with my pronouns, or why we were so busy trying to define what a woman is (isn’t it obvious? 😅) All of it was, and still is, a distraction from real problems.
0:59

One of the reasons why I left the USA was because of woke culture. I didn’t understand why I had to wear a name tag with my pronouns, or why we were so busy trying to define what a woman is (isn’t it obvious? 😅) All of it was, and still is, a distraction from real problems.

Sasha Meets Russia

67,769 Aufrufe • vor 1 Jahr

ARC-AGI-3 is built different, it has dumbfounded almost all regular attempts so far because it's so much harder than anything that came before. It has no rules, it's agentic and has no explicit goals, they need to be discovered. Tufalabs won the first milestone of ARC Prize > There is no language built into the benchmark, but these guys "put the language back in", because in their view - it's the best way to climb up the notional "abstraction mountain" and effectively use many of the abstractions which have evolved over millions of years of language evolution. > They built a novel harness "The Duck" around a 27B open weights model (Qwen 3.6) to solve extremely challenging and novel reasoning problems that require abstraction. > This is the launch video of their winning agentic harness, "The Duck". We have also released an exclusive interview with them on MLST, just dropped. > The million dollar question is: what will François Chollet think about how they've done it, and is this a step towards AGI?
33:13

ARC-AGI-3 is built different, it has dumbfounded almost all regular attempts so far because it's so much harder than anything that came before. It has no rules, it's agentic and has no explicit goals, they need to be discovered. Tufalabs won the first milestone of ARC Prize > There is no language built into the benchmark, but these guys "put the language back in", because in their view - it's the best way to climb up the notional "abstraction mountain" and effectively use many of the abstractions which have evolved over millions of years of language evolution. > They built a novel harness "The Duck" around a 27B open weights model (Qwen 3.6) to solve extremely challenging and novel reasoning problems that require abstraction. > This is the launch video of their winning agentic harness, "The Duck". We have also released an exclusive interview with them on MLST, just dropped. > The million dollar question is: what will François Chollet think about how they've done it, and is this a step towards AGI?

Machine Learning Street Talk

33,410 Aufrufe • vor 1 Monat

It was great catching up with Matt Barnes and Stephen Jackson on Home Court at the Obama Presidential Center. We spoke about the game we love, what it means to open the Center in Chicago, and why understanding our history is so important for navigating the world today. We had a lot of fun—check out our conversation on ALL THE SMOKE:
1:28

It was great catching up with Matt Barnes and Stephen Jackson on Home Court at the Obama Presidential Center. We spoke about the game we love, what it means to open the Center in Chicago, and why understanding our history is so important for navigating the world today. We had a lot of fun—check out our conversation on ALL THE SMOKE:

Barack Obama

6,655,069 Aufrufe • vor 1 Monat

Introducing GPT-Realtime-2 in the API: our most intelligent voice model yet, bringing GPT-5-class reasoning to voice agents. Voice agents are now real-time collaborators that can listen, reason, and solve complex problems as conversations unfold. Now available in the API alongside streaming models GPT-Realtime-Translate and GPT-Realtime-Whisper — a new set of audio capabilities for the next generation of voice interfaces.
4:03

Introducing GPT-Realtime-2 in the API: our most intelligent voice model yet, bringing GPT-5-class reasoning to voice agents. Voice agents are now real-time collaborators that can listen, reason, and solve complex problems as conversations unfold. Now available in the API alongside streaming models GPT-Realtime-Translate and GPT-Realtime-Whisper — a new set of audio capabilities for the next generation of voice interfaces.

OpenAI

3,654,751 Aufrufe • vor 2 Monaten

The Inverted Shoe👟 So the right shoe that was found at the Good Samaritan Hospital was placed in a bag that says it was the left shoe that was found at 34 Fairview…? Don’t worry guys, the evidence has not been tampered with or manipulated in any way!! We can save the hyperbole!
0:38

The Inverted Shoe👟 So the right shoe that was found at the Good Samaritan Hospital was placed in a bag that says it was the left shoe that was found at 34 Fairview…? Don’t worry guys, the evidence has not been tampered with or manipulated in any way!! We can save the hyperbole!

Look the Other Way

48,035 Aufrufe • vor 1 Jahr

#AusGP 🇦🇺 | Lewis Hamilton post qualifying: “The whole weekend was looking good up until Q2. Q1 on the medium tyre was feeling solid, and I was feeling great.” “Then we went into Q2 and we had some problems with our engine so we ended up having to come in [to the garage]. That put a lot of pressure on us to have to go out and try to execute with one lap, on a tyre that we hadn’t driven yet. That was tricky.” “Then we went to Q3, again, it was just mess for everybody, so it was a bit random. I think there’s a lot more performance in the car, we just didn’t executed those perfectly. I honestly think if it had gone perfectly we could’ve been 3rd today.”
1:01

#AusGP 🇦🇺 | Lewis Hamilton post qualifying: “The whole weekend was looking good up until Q2. Q1 on the medium tyre was feeling solid, and I was feeling great.” “Then we went into Q2 and we had some problems with our engine so we ended up having to come in [to the garage]. That put a lot of pressure on us to have to go out and try to execute with one lap, on a tyre that we hadn’t driven yet. That was tricky.” “Then we went to Q3, again, it was just mess for everybody, so it was a bit random. I think there’s a lot more performance in the car, we just didn’t executed those perfectly. I honestly think if it had gone perfectly we could’ve been 3rd today.”

deni

100,903 Aufrufe • vor 4 Monaten

And why was it that Israel closed the Allenby crossing with Jordan? Because of our innate cruelty? Or was it because of an aid driver carrying out a terror attack that killed two people? The crossing has now been reopened, so not only is this “journalistic” video lacking all the relevant context, it is also spreading a lie.
0:55

And why was it that Israel closed the Allenby crossing with Jordan? Because of our innate cruelty? Or was it because of an aid driver carrying out a terror attack that killed two people? The crossing has now been reopened, so not only is this “journalistic” video lacking all the relevant context, it is also spreading a lie.

Hen Mazzig

10,588 Aufrufe • vor 7 Monaten

GPT-5.6 Sol Ultra lost to GPT-5.5 on physics at 3× the cost! We gave 4 models the same prompt: build three self-contained HTML5 canvas scenes with real physics demos Prompts: - A monster truck backflipping onto a parked car - A stunt car jumping six buses into a brick wall - A train derailing off a broken bridge into the water Outputs: GPT-5.6 Sol Ultra: 32.9K tokens, $0.33 Opus 4.8: 9.2K tokens, $0.24 GPT-5.5: 12.4K tokens, $0.11 Grok 4.5: 7.0K tokens, $0.08 Sol Ultra draws just like GPT-5.5, only with more detail and it shows clearest on the bus jump where the two look almost identical. In the other tests Sol Ultra came out worse than its predecessor and the physics really got weaker. We think GPT-5.5 took the truck flip and the train outright. With Sol Ultra you basically get GPT-5.5 with weaker physics and a nicer picture for 3x the price. Newborn Grok 4.5 failed two of the three tests and only came good on the train
0:25

GPT-5.6 Sol Ultra lost to GPT-5.5 on physics at 3× the cost! We gave 4 models the same prompt: build three self-contained HTML5 canvas scenes with real physics demos Prompts: - A monster truck backflipping onto a parked car - A stunt car jumping six buses into a brick wall - A train derailing off a broken bridge into the water Outputs: GPT-5.6 Sol Ultra: 32.9K tokens, $0.33 Opus 4.8: 9.2K tokens, $0.24 GPT-5.5: 12.4K tokens, $0.11 Grok 4.5: 7.0K tokens, $0.08 Sol Ultra draws just like GPT-5.5, only with more detail and it shows clearest on the bus jump where the two look almost identical. In the other tests Sol Ultra came out worse than its predecessor and the physics really got weaker. We think GPT-5.5 took the truck flip and the train outright. With Sol Ultra you basically get GPT-5.5 with weaker physics and a nicer picture for 3x the price. Newborn Grok 4.5 failed two of the three tests and only came good on the train

atomic.chat

896,867 Aufrufe • vor 22 Tagen