Economist Tyler Cowen on How ChatGPT Is Changing Your Job - Ep. 7 with Tyler Cowen

[@tripp8833]

This podcast is awesome

[@eugeniocg3079]

incredibly informative

[@edism]

10 mins in so far, to hell with what I was doing earlier this is a great discussion already. Thank you for uploading. Shared!

[@johnexley.family]

Incredible intro, incredible opening question. So excited to watch this !

[@danshipper7738]

thanks John!

[@tomasflores_art]

Great Episode! Through the talk i've had the sense of seeing GPT more "humanized" in the way Tyler Cowen uses, even though the Playground is less user friendly.

[@shanecorreia]

I came here from a link in a newsletter, and I am spending a little bit too much time trying to figure out how to open this not on KZbin, but on a podcast app that I can safely listen to during my day without needing to keep the app open or click on something and lose my place

[@EveryInc]

Check us out on Spotify! open.spotify.com/show/5qX1nRTaFsfWdmdj5JWO1G?si=xZVHwPikQt6ZKvcyFiahpg

[@deanwu6748]

🎯 Key Takeaways for quick navigation: 00:00 🌐 *Economist Tyler Cowen discusses the impact of ChatGPT on productivity, seeing AI as a tool for learning and references.* 02:39 📚 *Tyler Cowen's book "Average Is Over" (2013) predicted the stratification in the economy due to intelligent machines, particularly AI, impacting different skill sets.* 04:45 🤖 *Cowen foresaw AI's potential breakthrough before 2013, driven by his belief in its capabilities, especially after witnessing AI advancements in chess.* 07:20 💼 *The short-term impact of AI is expected to be egalitarian, helping those with limited skills to accomplish tasks they couldn't do before, but the longer-term effects might favor those initiating and managing projects.* 08:42 🔄 *The allocation economy may lead to more widespread distribution of managerial skills as AI facilitates resource allocation, potentially making management advice more accessible.* 11:28 🧠 *Cowen discusses the increasing importance of skills related to allocation and management in the evolving economy, with AI potentially automating certain aspects of managerial advice.* 13:19 ⚠️ *Cowen highlights an underappreciated risk of AI in aiding nefarious activities, where improved management advice for terror organizations could pose a significant threat.* 15:17 🔄 *Cowen envisions a short to medium-term scenario where AI accelerates the value of skills like ordering information and knowledge, but the long-term may see a decline in the value of these skills.* 16:51 📱 *Cowen shares two distinct ways he uses ChatGPT: as a universal translator and for reading menus while traveling, enhancing the interactive discovery experience with others.* 18:56 🌍 *The use of ChatGPT in travel adds flavor to the trip by providing an interactive experience with others when discovering and learning new things.* 19:50 🤔 *Tyler Cowen compares ChatGPT to a magical information tool, highlighting its quick and efficient retrieval of facts, even in remote locations.* 21:12 🔄 *Tyler discusses his approach to checking ChatGPT's output for correctness, emphasizing the low concern for hallucinations and the effectiveness of asking it to correct errors.* 23:40 📚 *Tyler shares his positive experience using ChatGPT to understand and visualize passages from old books like Moby Dick, leveraging the model's access to extensive text and supporting commentary.* 27:51 💡 *Tyler explains his preference for the ChatGPT playground, citing its control over output parameters, system messages, and content restrictions, despite acknowledging potential trade-offs.* 29:56 🧠 *Tyler uses ChatGPT to inquire about the historical development of the distinction between whole life and term life insurance, highlighting the model's role in spurring thoughts and providing additional information.* 34:54 🔄 *Tyler explores the value of adjusting ChatGPT's temperature for more creative responses, showcasing the potential for nuanced and prioritized answers in response to specific queries.* 36:21 🎭 *The experiment with temperature in ChatGPT's responses underscores the playground's exploration capabilities, allowing users to find more creative or refined answers.* 38:51 ❓ *Tyler's attempt to extract specific inflation figures from ChatGPT for the Byzantine Empire encounters limitations in the model's commitment to providing exact numbers, revealing certain constraints in its responses.* 39:57 🧠 *Tyler Cowen emphasizes the importance of managers' skills in assessing information credibility, knowing when to check and follow up on details.* 41:08 🌐 *Tyler Cowen discusses his information gathering process, starting with ChatGPT, then moving to perplexity AI, checking references, and finally reaching out to people or using Google for links.* 43:50 🤖 *Tyler Cowen distinguishes the roles of Google (for links), AI (for learning), and perplexity (for references) in his information-seeking strategy.* 46:13 📚 *Tyler Cowen shares his experiment using GPT for book recommendations and highlights its effectiveness in identifying personal taste and suggesting new, undiscovered books.* 50:18 🎓 *Tyler Cowen discusses using ChatGPT in the classroom, making it a requirement for students in his history of economic thought class, encouraging its use for writing papers while emphasizing responsible usage.* 51:52 🌐 *Foreign students, especially from China, use ChatGPT to improve their English writing skills significantly, with transformative effects.* 53:56 🔄 *Tyler Cowen discusses the translation aspect of GPT, mentioning how it can help individuals translate their ideas into a more sophisticated form, such as turning non-technical language into a product spec.* 55:19 🤯 *Tyler Cowen addresses the potential impact of ChatGPT's absence on his productivity, highlighting the nuanced relationship between productivity, intelligence, and the value of being smart in labor markets.* 56:26 🌐 *Tyler Cowen talks about the trade-off between using ChatGPT and maintaining a reading flow, acknowledging that it slows down the reading process but can be valuable for learning in unfamiliar areas.* 57:46 🗣️ *Tyler Cowen expresses openness to the idea of using an AI clone for text-to-speech purposes, citing an example of an AI reading the audiobook version of his book.* 59:38 🤖 *Tyler introduces "Tyler bot" and evaluates its responses.* 59:52 🧠 *Core lessons of Economics: Incentives matter and there's always an opportunity cost.* 01:00:12 🎓 *Tyler rates Tyler bot's performance, emphasizing the value of brevity in economics.* 01:00:42 🤵 *Describing Tyler: Curious, analytical, eccentric, and intellectual.* 01:03:24 🤖 *AI's impact on economics: Tyler expects no fundamental change in the next decade, envisions simulating small economies in the future.* 01:04:37 🤖 *Tyler bot's response: Predicts radical transformation in information costs and decision-making processes, a generic answer according to Tyler.* 01:05:33 📚 *Tyler recommends his book "Goat. a" and highlights its unique feature of being embedded in an AI chatbot, suggesting this could be the future of some book publishing.* Made with HARPA AI

[@megamalone79]

How did you create TylerBot?

[@MarkMyerson]

I definitely think society is going to face major issues with regard to relying on AI for factual knowledge. ChatGPT might seem more balanced than many traditional sources of information. However, this won't necessarily apply always. And it almost certainly won't apply to every model. I can imagine a future where AI providers fulfill the same roles as Fox News and CNN - each potentially delivering a potently lopsided and emotive take on reality. And because people won't search Google for information, they won't even have to scroll past dissenting headlines. I also wonder how Tyler's view on accuracy would scale. Even if he has a good understanding of when hallucinations matter and when they don't, I wonder whether the average user is going to be so discerning. And that includes decision-makers who shape our world.

[@jamiewalden7774]

It can't possibly be worse than a large portion of society's existing reliance on facebook for "factual" knowledge.

[@davdfranzen]

Perplexity uses Google search results