Artificial intelligence is transforming the business world in ways we couldn't have imagined until recently. Just how--and what the future holds--was the topic of a panel discussion at the TIME100 Impact Dinner: Leaders Shaping the Future of AI in San Francisco on Monday moderated by TIME's executive editor Nikhil Kumar. The panelists were Ravi Kumar S, CEO of Cognizant, which sponsored the event; Athina Kanioura, chief strategy and transformation officer at PepsiCo, which also sponsored the event; and Jared Kaplan, co-founder and chief science officer at Anthropic . Ravi Kumar and Kaplan both featured on the 2025 TIME100 AI list, which highlights the 100 most influential people in AI this year, from computer scientists to business leaders to policy makers and artists. "One of the things about the public conversation about AI is that quite often it is focused on the companies behind the technology and what they are doing," said TIME's Nikhil Kumar when introducing the panel.

Citizens' assemblies are an increasingly influential form of deliberative democracy, where randomly selected people discuss policy questions. The legitimacy of these assemblies hinges on their representation of the broader population, but participant dropout often leads to an unbalanced composition. In practice, dropouts are replaced by preselected alternates, but existing methods do not address how to choose these alternates. To address this gap, we introduce an optimization framework for alternate selection. Our algorithmic approach, which leverages learning-theoretic machinery, estimates dropout probabilities using historical data and selects alternates to minimize expected misrepresentation. Our theoretical bounds provide guarantees on sample complexity (with implications for computational efficiency) and on loss due to dropout probability mis-estimation. Empirical evaluation using real-world data demonstrates that, compared to the status quo, our method significantly improves representation while requiring fewer alternates.

Link functions, which characterize how human preferences are generated from the value function of an RL problem, are a crucial component in designing RLHF algorithms. Almost all RLHF algorithms, including state-of-the-art ones in empirical studies such as DPO and PPO, assume the link function is known to the agent (e.g., a logistic function according to the Bradley-Terry model), which is arguably unrealistic considering the complex nature of human preferences. To avoid link function mis-specification, this paper studies general RLHF problems with unknown link functions. We propose a novel policy optimization algorithm called ZSPO based on a new zeroth-order policy optimization method, where the key is to use human preference to construct a parameter update direction that is positively correlated with the true policy gradient direction. ZSPO achieves it by estimating the sign of the value function difference instead of estimating the gradient from the value function difference, so it does not require knowing the link function. Under mild conditions, ZSPO converges to a stationary policy with a polynomial convergence rate depending on the number of policy iterations and trajectories per iteration. Numerical results also show the superiority of ZSPO under link function mismatch.

Russian state media is now mocking former president Donald Trump after he was indicted on 34 felony counts of falsifying business records for charges stemming from an alleged hush money payment to an adult film star. Last Friday, the Russian state show "60 Minutes" displayed an AI-generate image of Trump wearing an orange outfit while panelists discussed what could happen in the United States should the former president be arrested. Meanwhile on Russian state TV, even the biggest fans of the former president are using AI-generated images of Trump in orange. Another picture of Trump wearing similar orange overalls was also used on "Sunday Evening with Vladimir Solovyov." During the weekend broadcast, host Vladimir Solovyov and the show's panelists wondered if Trump has a chance of winning the 2024 election following the indictment.

The Twelve Labs team had a blast hosting its first Dream in Vectors fireside chat in Seoul where we invited a group of passionate panelists, ML research scientists, engineers, and students to reflect on 2022 and dream about what's to come in 2023. The sheer amount of knowledge from the group was immeasurably grand and the future of AI is so so so bright! There were many highlights and to share a few topics that we discussed: 1 Generative search with a strong grounding in factual references will disrupt search from the ground up. 2 Multimodal is the clear next step and we will see very large visual language models next year.

The panel event will be part of a series of workshops conducted over two days at the conference. Organisers will record each of the workshops and feed the text transcripts through OpenAI's ChatGPT and Microsoft's Bing AI to create a report of the two-day event. ChatGPT and Bing AI will also join the discussion and answer audience questions alongside human panellists. The two AI models have been ridiculed for giving false answers to search queries, but new iterations continue to improve on the nuances of human thinking. OpenAI, for instance, has just released the latest version of ChatGPT – ChatGPT-4 – which it says is more creative, less likely to make up facts and less biased than its predecessor.

Since "google" entered our vocabulary as a synonym for "search," Google's dominance in the search market has seemed unshakable, making the very idea of competition between search engines seem unlikely. But the entrance of generative AI tools like ChatGPT -- developed by OpenAI in partnership with Microsoft -- has ignited the potential for a competitor to knock Google's capabilities off the top of the search pyramid. Microsoft's new Bing Chat, which integrates the same AI technology that powers the ChatGPT bot, is already generating curiosity, controversy, and a long waitlist of would-be users. Google's parent company, Alphabet, has also developed its own AI chatbot, Bard. The future success and impact of these new technologies are very much open questions, so we turned to our expert panelists for their responses to this statement: The use of generative AI will restore competition in search.

Late last month, the Biden Administration announced that the U.S. would send thirty-one M1 Abrams tanks to Ukraine. Meanwhile, in New York, a Ukrainian delegation, including a representative of the Territorial Defense Forces, had gathered to consider other types of aid. The goal, according to an ad for the event, was to promote "the psychological and spiritual resilience of Ukrainian people living in trauma, crisis, and war." The delegation met at a studio in Chelsea run by a Polish artist named Agnieszka Pilat. She paints with the aid of mobile robots on loan from Boston Dynamics; a yellow robot that resembled a dog pattered around the space as the audience arrived.

MIT Sloan Management Review and BCG have assembled an international panel of AI experts that includes academics and practitioners to help us gain insights into how responsible artificial intelligence (RAI) is being implemented in organizations worldwide. This month's question for our panelists: Should RAI be a top management agenda item at organizations across industries and geographies?1 Eighty-six percent of them (18 out of 21) agree or strongly agree that it should be. In aggregate, their replies offer a compelling rationale for top management to oversee RAI efforts. We distill and explain this rationale below. We also conducted a global survey of more than 1,000 executives that generated similar findings: Eighty-two percent of managers in companies with at least $100 million in annual revenues agree or strongly agree that RAI should be part of their company's top management agenda.

I Can't Believe It's Not Better: Understanding Deep Learning Through Empirical Falsification Workshop Organizers include: Javier Antorán Panelists include: Kevin Murphy