Patreon.com/svic
Sully discusses the creation and development of Cognosys, an AI platform that automates workflows using natural language processing, including the challenges faced and solutions implemented. They explore open problems in AI development, such as billing and termination issues. The conversation delves into the evolution of AI models, the future of retrieval augmentation generation (RAG) and large context models, and topics like simplifying AI systems, agentic models, and surprising use cases.
Show hosts Joe (Eng VP) and Jordan (M&A Deal Lead) worked at various companies from Google, Apple, Facebook, Microsoft, Salesforce, Slack, Carta, Splunk, Wealthfront, Adobe, Microsoft, and more.
????Here the rest of the interview & get access to new stories and blog posts for the episode and the ones we didn’t have time to cover, our book recommendations, exclusive episodes and latest AI research for only $5 at patreon.com/svic
Takeaways
00:00 Introduction and Background
01:35 How Cognosys.ai went virtal
05:30 How AutoGen Evolved Into Cognosys
12:17 Picking The Right LLMs
18:00 Big Challenges Cognosys Is Trying To Solve
20:43 How GPT-5 Can Make Your AI Technique Obsolete
31:20 Determine Product Pricing
33:45 Customer Surprises And Product Roadmap
37:45 GPT-4 Dethroned By Claude 3
42:45 Using Claude 3 to write code
#openai #claude #ai #artificialintelligence #chatgpt #gpt4
