Chris
@ChrissGPT • 52,574 subscribers
Agi 2029 - AI Insider / Reporter as featured in The Information • NYT • Techcrunch
Shorts
Videos
Opus 4.7 cs GPT 5.4 “make a flappy bird clone” I’ve been running this test for a few models now and this test was the hardest to objectively say who did better. Opus grabbed another version game assets compared to GPT 5.4, - GPT 5.4 however, grabbed the correct game assets. Opus 4.7 beat it on creativity where it grabbed assets (maybe unknowingly) that had a much more diverse set, including nighttime modes and different bird colors GPT 5.4 was able to get the pacing of the game the first shot, whereas Opus had the game running much too fast and then much too slow. But was able to place the metal perfectly in the game card at the end where GPT 5.4 struggled over 10 iterations. All in all both models had essentially made a near perfect game after three shots and are far ahead of the competition in terms of reasoning and effort.
Chris423,821 次观看 • 6 个月前
🚨 SAM ALTMAN: "A REDUCTION IN COST OF ABOUT A 1000X" 🤯 Sam Altman revealed the cost to solve a hard reasoning problem has plummeted by 1,000x since the release of o1 just 16 months ago (comparing o1 to the new 5.4 model). How is this possible? Altman credits two main factors: 1️⃣ We're Still in the Early Days: “We are so new to this paradigm that there is massive room for improvement. We're still doing things in "dumb ways" and getting better fast.” 2️⃣ Human Ingenuity is Underrated: “It's not just the models getting better. It's kernel engineers, power engineers, and data center designers finding brilliant ways to optimize every step of the process.” The compounding effect of better models + better human engineering is driving an exponential decrease in the cost of AI reasoning.
Chris283,126 次观看 • 7 个月前
I don’t know how they’re getting the quality so damn close to the actual show simply by training off of it. I remember posing this question to my professor, if diffusion models would ever be able to get one to one quality simply by training off the pixels and then manipulate it as if they actually understood the scene rather than just parroting. Back then it felt almost impossible. The assumption was that you would hit a ceiling because the model was “just” learning statistical correlations in pixel space. No semantic grounding. No true understanding. It almost feels like seedance is like an alien that can go beyond noise to replicate 1 to 1 quality sometimes.
Chris130,109 次观看 • 7 个月前
SAM ALTMAN: "You will graduate to a world with AGI in it" Sam Altman just told a room of college sophomores that by the time they finish school, Artificial General Intelligence will be a reality. Specifically: • Science will become increasingly automated. • What it means to start a startup or work at a company will be totally different. • A lot of traditional career advice is no longer going to work. Which is all true. My AGI timeline is actually slightly later than Sam’s - Whereas I believe it will arrive in 2029 but at that point we are grasping at straws.
Chris108,244 次观看 • 7 个月前
Google Deepmind CEO gives his prediction on the final AGI architecture: Demis Hassabis just sat down on 20VC and broke down exactly how DeepMind is building toward true AGI, directly pushing back on the idea that LLMs are a dead end. He defends the current trajectory of the models, noting that “we’ve seen how successful these foundation models have been. They can do incredibly impressive things. I don’t think that’s going to go away. We’re still seeing returns from the scaling laws.” But he’s honest about the missing pieces for true superintelligence, saying “I think there might be… there’s a 50-50 chance there’s some things maybe missing that we still need to make breakthroughs in, perhaps their world models, these kinds of approaches.” And when he breaks down what the actual AGI system will look like, he frames it as one question is the LLM foundation model the key component only, or is it the total system. His answer - “I don’t think it’s going to get replaced. I think it’s going to get built on top of these foundation models, just like the way we do with our world models.”
Chris71,716 次观看 • 5 个月前
0:52
Sensitive content
This media may contain sensitive content.
I’ve been seeing a lot of slop games made with Claude Fable 5, so I wanted to build something I’d actually enjoy playing. What better way to celebrate all the new robotics companies shipping this year than a first-person robot simulator where you clean and organize a house? This was mostly vibe prompted into a small robotics sandbox, but it gave me a whole new appreciation for how difficult robotics actually is. Even in a game, controlling two independent arms in 3D space, lining up grasps, avoiding collisions, and making sure your own body doesn’t knock things over is surprisingly challenging 😆, It looks simple until you play. Definitely one of the most fun projects I’ve made so far, and I want to keep expanding it. Still on the slop indie stage but the majority people watching this video will quickly forget how creating anything like this was genuinely impossible year ago and I implore you to try to one shot something like this with the models available a year ago.
Chris32,615 次观看 • 3 个月前
My final hours with Claude Fable 5 My Kerbal Space Program inspired game In honor of the SpaceXai IPO 😄 3 shot - 67k lines of code I realized I can make this into a ballistic missile simulator. One thing I like is you can make the rocket absurdly big and add as many engines as you want. But if you watch toward the end, my favorite feature I added was the decoupling stage
Chris22,172 次观看 • 4 个月前
没有更多内容可加载