Saturday, July 13, 2024

This Week in AI with Reed Hepler and Steve Hargadon (July 12, 2024)

We've released our most recent "This Week in AI" recording. Hope you enjoy! AI summary provide by summarize.tech: https://www.summarize.tech/www.youtube.com/watch?v=EjjqA3XpaCI.

In the July 12, 2024 episode of "This Week in AI," hosts Steve Hargadon and Reed Hepler discuss their personal experiences with AI's role in productivity and share news about the latest developments in the field. They reflect on the shift towards using AI for productive purposes and the growing trend among librarians to use AI for conversational search. In the news segment, they cover a study questioning the data analyzing abilities of Google's Gemini AI and the importance of verifying the data processing powers of AI tools. The hosts also discuss the use of AI chatbots in schools, the potential consequences of relying on AI technology, and the role of AI in fact-checking and medicine. They express excitement about the potential of AI in diagnosing diseases and creating proteins for drugs but raise concerns about peer review, patent ownership, and regulatory responses. Throughout the conversation, they emphasize the importance of adapting to the rapidly changing technological landscape and focusing on the techniques behind the tools rather than the tools themselves.

  • 00:00:00 In this section of "This Week in AI" from July 12, 2024, hosts Steve Hargadon and Reed Hepler discuss their personal experiences with AI's role in productivity and share news about the latest developments in the field. Both hosts have noticed a shift towards using AI for productive purposes rather than just for fun. They also mention the growing trend among librarians to use AI for conversational search instead of database search. In the news segment, they cover a study revealing that Google's Gemini AI may not live up to its claimed data analyzing abilities, and the increasing importance of verifying the data processing powers of AI tools as companies make bold promises. The hosts acknowledge the pressure on companies to present a positive image and the potential for overpromising.
  • 00:05:00 In this section of the "This Week in AI" video from July 12, 2024, Steve Hargadon and Reed Hepler discuss the limitations and potential of large language models in handling logic and data analysis. Hepler explains that the misconception lies in the belief that AI can perform logic independently, while humans must provide context and control the conversation. They also touch upon the development of AI products like "eternal you," which can create a likeness of a person and bring comfort to users. Hepler expresses skepticism about the desire for such technology, while Hargadon sees it as a potential emotional fulfillment tool. The conversation then shifts to the Turing test and the idea that AI may surpass human perception and intellectual capacity, leading to questions about the impact on human skills and incentives to think deeply.
  • 00:10:00 In this section of the "This Week in AI" video from July 12, 2024, Reed Hepler and Steve Hargadon discuss the use of AI chatbots in schools, specifically mentioning the Los Angeles school district. The chatbot was intended to provide information to parents and students about grades, homework, and district news. However, the company behind the chatbot went bankrupt, and there were concerns about compromised data. Hepler expresses his belief that humans are necessary for handling sensitive information related to children. Steve Hargadon adds that Los Angeles has a history of spending large sums of money on technology that fails to improve student outcomes. The conversation then shifts to the potential of AI in improving government services, with Tony Blair, a former Prime Minister, advocating for its use in a recent report. Despite the potential benefits, both speakers caution against over-reliance on AI and the importance of careful consideration before implementation.
  • 00:15:00 In this section of the "This Week in AI" video from July 12, 2024, Steve Hargadon and Reed Hepler discuss the rush to implement AI technology and the potential consequences, particularly in the context of Google searches. The speakers express concern over the exaggeration of AI capabilities and the impact on industries like content creation. They also discuss the decline in mobile searches due to AI overviews providing answers directly, causing people to bypass visiting websites. This trend raises concerns about the death of expertise and the potential loss of revenue for content creators. Despite the flaws in the current AI overviews, the convenience of having answers delivered instantly is a significant draw for users.
  • 00:20:00 In this section of the "This Week in AI" video from July 12, 2024, Reed Hepler and Steve Hargadon discuss the reliance on AI tools and the importance of accuracy. Hepler shares his concern about the use of AI to assess the accuracy of AI summaries, creating a dependency on multiple AI systems. Hargadon adds to the conversation, reflecting on human behavior and the tendency to choose ease over responsibility. He argues that this trend will continue with AI, as people may prioritize convenience over accuracy or long-term impact. The conversation touches on the Amish theme of evaluating technology's impact and the potential need for legislation to address these issues.
  • 00:25:00 In this section, Steve Hargadon discusses the potential of AI in fact-checking and improving its own capabilities. He shares his experiment of creating a custom GPT model preloaded with books on historical misrepresentations to evaluate news stories with skepticism. The AI performed well in responding to specific stories, leading Hargadon to believe that future iterations could help improve accuracy. Reed Hepler adds that many AI models, including OpenAI and Clot, use models to train other models, creating a reinforcement learning process without human intervention. The conversation then shifts to the intersection of AI and genetics, with Reed Hepler highlighting the recent development of a gigantic AI protein design model by X-Meta, which aims to identify and create proteins relevant to disease research and mutation research. The model, created by Evolutionary Scale, has been made compatible with Crispr technology and is expected to change the field of medicine and drug development. Additionally, AI is being used to analyze the genetics and DNA of tumors to understand their functioning and response to treatment.
  • 00:30:00 In this section of the "This Week in AI" YouTube video from July 12, 2024, Reed Hepler and Steve Hargadon discuss the potential impact of artificial intelligence (AI) on various industries, specifically medicine and pharmaceuticals. They express excitement about the potential for AI to diagnose diseases and create proteins for drugs, but also raise questions about peer review, patent ownership, and regulatory responses. Hargadon also emphasizes the importance of adapting to the rapidly changing technological landscape and focusing on the techniques behind the tools rather than the tools themselves. Hepler agrees, and they both look forward to exploring these topics further.

No comments:

Post a Comment

I hate having to moderate comments, but have to do so because of spam... :(