Technology

Experts urge caution about using ChatGPT to pick stocks

Source

Ars Technica

Published

Sep 25, 2025

TL;DR

AI Generated

A growing number of retail investors are turning to AI chatbots like ChatGPT for stock-picking advice, with around 1 in 10 investors consulting these tools, according to a Reuters report. Data from a survey of 11,000 retail investors worldwide shows that 13 percent are already using AI tools like ChatGPT or Google's Gemini for stock selection, and half are open to using them for portfolio decisions. Unlike algorithmic trading, where computers execute trades automatically, investors are using ChatGPT as an advisory tool, typing questions and manually deciding whether to act on the AI model's analysis through their brokers. Experts caution about relying solely on AI chatbots for stock market decisions.

Read Full Article

South Korean company following ChatGPT management advice forced to backtrack by court — Krafton improperly ousted Subnautica developer bosses to dodge $250m bonus payout

South Korean company Krafton Inc. has been ordered by a Delaware judge to reinstate the leadership of U.S. game studio Unknown Worlds Entertainment, which it acquired in 2021. Krafton improperly ousted the Subnautica developer bosses to avoid a $250 million bonus payout. The CEO of Krafton sought advice from ChatGPT to navigate the situation, leading to the removal of the studio's founders and CEO without proper cause. The court ruled in favor of the U.S. studio leadership, extending the earnout period for the bonus and reinstating the CEO. Krafton disagrees with the ruling and is exploring its options.

Tom's Hardware•

2 months ago

MIT Technology Review

“Dr. Google” had its issues. Can ChatGPT Health do better?

OpenAI's ChatGPT Health is positioned as a potential improvement over relying on "Dr. Google" for medical advice. With 230 million health-related queries per week, ChatGPT Health aims to provide guidance and tools for health advice, emphasizing it as a supplement to professional medical care. While large language models like ChatGPT can help navigate medical information and improve medical literacy, concerns remain about their accuracy and potential to spread misinformation. Studies show that LLMs can answer medical questions correctly a high percentage of the time, but there are limitations in evaluating their effectiveness for complex health issues. Despite advancements in AI models, caution is advised in relying solely on AI for medical guidance over human doctors.

MIT Technology Review•

4 months ago

Famed gamer creates working 5 million parameter ChatGPT AI model in Minecraft, made with 439 million blocks — AI trained to hold conversations, working model runs inference in the game

Famed gamer Sammyuri has created CraftGPT, a working 5-million parameter ChatGPT AI model in Minecraft using 439 million blocks. The AI model is trained to hold conversations and runs inference within the game, but response generation can take hours. CraftGPT is built using Redstone components in Minecraft without command blocks or data packs. Despite its limitations in chat quality and performance, CraftGPT showcases an impressive technical achievement within the Minecraft environment.

Tom's Hardware•

8 months ago

ChatGPT Pulse delivers morning updates based on your chat history

OpenAI has introduced ChatGPT Pulse, a new feature that provides personalized daily updates to users without requiring them to ask each time. This feature, available for Pro subscribers on mobile, aims to make ChatGPT more proactive by using AI to conduct overnight research and deliver morning updates based on user history and connected apps. ChatGPT Pulse analyzes chat history, saved preferences, and optional connections to Gmail and Google Calendar each night to generate visual "cards" with topic summaries like project follow-ups or travel recommendations. Users can offer feedback through ratings and request specific topics, and updates disappear after 24 hours unless saved or followed up on, converting them into standard chat conversations.

Ars Technica•