Our paper on LTM Benchmark got accepted to NeurIPS 2024! Proud to be part of this talented team. Looks like our collective effort will be remembered for a while. Our contribution: We go beyond just info retrieval in long-context benchmarks. Our approach evaluates how LTM
@marek_rosa
-
AI Agents Will Widen Wealth Gap Between Rich Others
By
–
AI will widen the gap between the very rich and others, contrary to the original tweet. Example from game dev: Today: Billionaires and everyday people play identical games (see Elon Musk streaming Diablo) Future: The wealthy will use millions of AI agents vs 1 AI agent for
-
Demis Hassabis Wins Nobel Prize in Chemistry for AlphaFold 2
By
–
Big congratulations to Demis Hassabis on winning the 2024 Nobel Prize in Chemistry for AlphaFold 2! Your groundbreaking work at the intersection of AI and chemistry is truly inspiring. This recognition highlights how AI research is revolutionizing multiple scientific fields.
-
AI People Now Support Voice Interaction Feature
By
–
New update: You can now talk to our AI People using your voice! Try it out – it truly transforms the experience. No more typing required!
-
LLM Quality Thresholds: GPT-4o-mini and Local Alternatives
By
–
There's some lower threshold of quality an LLM needs to have in order to be useful for our game. GPT4o-mini is such an example, and it has good pricing, but it's still cloud based. However there are other candidates for local LLM that we are investigating. LLMs are getting
-
Expanding NPC Interaction Capabilities Through Game Engine Design
By
–
Our NPCs don't need to be "programmed" for these specific interactions, but the game needs to allow that interaction. So for example, if the game doesn't allow NPC to pick up and move chair, then the NPC simply can't do it. For these reasons we are expanding the range of
-
GPT-4 Outperforms Humans in Creativity and Clinical Reasoning
By
–
My bet is that in long run, AI will be beter at creativity than humans, including entrepreneurship. Related to this is this study that says: GPT-4 achieves 88% diagnostic accuracy, outperforming doctors by 15% in clinical reasoning test Another intepretation: GPT-4 alone was
-
AI Agent Proliferation Drives Energy Demand Explosion
By
–
I meant mainly energy for inference, not training. Even if each AI agent consumed little (say, 1 iPhone's worth), the demand will explode as people want more agents – 2, then 10, 100, millions per person. It's this proliferation, not individual inefficiency, that'll drive energy
-
AI People Alpha Release Showcased by Idnes.cz Game Coverage
By
–
http://
Idnes.cz recently wrote about our new game AI People's alpha release. Their article offers some valuable perspectives: "AI People is one of the innovators that could pave the way for more believable and interesting artificial intelligence in games." "The game is -
Automating Founders: Can AI Entrepreneurs Replace Humans?
By
–
What do you think about automating the founder / replacing them with AI entrepreneur agent? Then the only economic input is capital (and GPUs/electricity for inference)