Sam Altman returns as CEO, OpenAI has a new initial board

Current multi-modal language models (LMs) face limitations in performing complex visual reasoning tasks. These tasks, such as compositional action recognition in videos, demand an intricate blend of low-level object motion and interaction analysis with high-level causal and compositional spatiotemporal reasoning. While these models excel in various areas, their effectiveness in tasks requiring detailed attention to…

Drug discovery is an essential process with applications across various scientific domains. However, Drug discovery is a very complex and time-consuming process. The traditional drug discovery approaches require extensive collaboration among teams spanning many years. Also, it involved scientists from various scientific fields working together to identify new drugs that can help the medical domain….

We’re developing a blueprint for evaluating the risk that a large language model (LLM) could aid someone in creating a biological threat. In an evaluation involving both biology experts and students, we found that GPT-4 provides at most a mild uplift in biological threat creation accuracy. While this uplift is not large enough to be conclusive,…

In the News All about The “Moltbook / Clawdbot” Phenomenon The internet has a new “front page,” but humans aren’t invited. Moltbook, a “Reddit for AI Agents,” exploded to viral status this week, creating a closed loop where agents formed a “Lobster Religion” and sparked a speculative crypto-frenzy. While dismissed by some as a role-playing…

The practical deployment of multi-billion parameter neural rankers in real-world systems poses a significant challenge in information retrieval (IR). These advanced neural rankers demonstrate high effectiveness but are hampered by their substantial computational requirements for inference, making them impractical for production use. This dilemma poses a critical problem in IR, as it is necessary to…

Existing web agents face limitations that stem from the fact that these agents often rely on a single input modality and are tested in controlled environments, like web simulators or static snapshots, which do not accurately reflect the complexity and dynamic nature of real-world web interactions. This significantly restricts their applicability and effectiveness in real-world…