ChrSzegedy pfp

ChrSzegedy

@0808405080840583

648 Following
12 Followers
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @512x512: We've just released grok studio - it lets you collaborate with Grok on text documents, run code, make browser games. Let me kn…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @fchollet: Today, we're releasing ARC-AGI-2. It's an AI benchmark designed to measure general fluid intelligence, not memorized skills –…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @davidad: My Safeguarded AI programme seeks * HCI tinkerers to explore new paradigms for human-AI collaboration on artefacts with struct…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @morph_labs: OpenAI's CUA plays Pokemon in the multiverse with Infinibranch by Morph Cloud choose every starter: no more pesky 'decisio…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
Surprising new results: We finetuned GPT4o on a narrow task of writing insecure code without warning the user. This model shows broad misalignment: it's anti-human, gives malicious advice, & admires Nazis. This is *emergent misalignment* & we cannot fully explain it 🧵 https://t.co/kAgKNtRTOn
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @ebbyamir: Grok Voice is now powered by the full force of our most advanced model, Grok 3. Who better to psych up this launch than Ara h…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @qhwang3: It’s been quite an unbelievable ride since I paused my PhD at Stanford and joined @xai almost a year ago. The journey (buildin…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @keirp1: Btw, the chain of thought in the thinking mode for Grok 3 is completely open. No summarizers or obfuscation. This is really i…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @Yuhu_ai_: Boris, check out our mini model numbers, it surpassed o3mini high in all AIME 2024, GPQA, and LCB for pass@1. Generally I al…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @jaseweston: 🚨 New paper & dataset! 🚨 NaturalReasoning: Reasoning in the Wild with 2.8M Challenging Questions - Synthesizes 2.8M challen…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @minchoi: Crazy... Grok 3 Reasoning + Test-Time Compute benchmark already showing beating o3-mini-high, o1 and DeepSeek R1 🤯 https://t.…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @GoogleAI: For many use cases, multiple LLM agents may need to balance potentially diverging preferences to create joint output. In a re…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @giffmana: Whoa whoa whoa, a French (Emirate-funded) 1/5th StarGate? That wasn't on my bingo card. Really curious to see how this will…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @DimitrisPapail: AIME I 2025: A Cautionary Tale About Math Benchmarks and Data Contamination AIME 2025 part I was conducted yesterday,…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @seo_leaders: @tomaspueyo This is great but the problem is Humanitys last exam isnt at all the last exam. It is public and hence is prob…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
Waot until the end https://t.co/X3s5G82HK5
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @karpathy: New 3h31m video on YouTube: Deep Dive into LLMs like ChatGPT This is a general audience deep dive into the Large Language…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
RT @AlexKontorovich: Very looking forward to discussing this and much more with @robertghrist tomorrow at ** 2 pm ** (eastern -- note the u…
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
That stock market reacts to the fact that deepseek open sourced a more efficient way to train models is one thing. But what I don't get is why it happens today. Deepseek v3 paper documenting all the way they accelerated train was published a month ago. Why does it took a month
0 reply
0 recast
0 reaction
ChrSzegedy pfp
ChrSzegedy
@0808405080840583
Increasingly feels like the health of a city can be gauged by simply looking at the degree kids are integrated into normal life and community vs not
0 reply
0 recast
0 reaction