AI-podden

Merry Christmas folks! This week, guest host (Sifted's) Mimi Billing and Ather discuss December's AI news updates. Highlights included Uber's entry into AI data labeling, Google DeepMind's Genie 2 for training AI in simulated environments, and Amazon's collaboration with Anthropic to develop a mega AI supercomputer. We covered the importance of new algorithms over mere scaling and the potential of agentic AI to bring advancements toward artificial general intelligence (AGI). The year saw record AI funding rounds, with Databricks leading at $10 billion, while disappointments included unmet expectations for reasoning capabilities in AI models. Looking ahead, we anticipate innovations in AI model architectures, memory integration, and task automation tools.

This week, guest host (Sifted's) Mimi Billing and Ather discuss November's AI news updates. Topics include; OpenAI’s Slush presentation, which revealed both innovation and cultural challenges, and the diminishing returns of AI scaling laws. They explore the rise of specialized models, new players like Gemesis and Mistral, and OpenAI's ChatGPT search feature. The episode also examines AI’s growing role in medicine, highlighting its diagnostic potential alongside concerns about biases and reliability, while reflecting on the balance between rapid progress and persistent challenges in AI development.

Christian Guttmann, Executive Director of the Nordic AI Institute, discusses AI as both a philosophical aspiration and a practical tool. He highlights applications like personalized banking recommendations and data querying with large language models but notes their limitations in reasoning and understanding. Advocating for hybrid AI models and richer data, he emphasizes the need for multimodal approaches that integrate various forms of input, such as speech and gestures, to advance AI capabilities. He also warns that Europe risks falling behind in AI development without greater investment and bold innovation.  

This week we welcome back; Reggie Townsend, Vice President of Data Ethics at SAS Institute and member of President Biden's National AI Advisory Council. We discussed AI development and regulation in the US. He stressed balancing innovation incentives with government restraint and cited the CHIPS Act as key for AI infrastructure, though most investments are private. He highlighted that AI trustworthiness varies, suggesting risk-based assessments. Townsend underscored the importance of AI education and common standards. On regulation, he noted a risk-focused, citizen-protective approach with inevitable dissent. Lastly, he urged AI to remain a bipartisan issue, avoiding politicization similar to electricity, with current bipartisan support.