ChatGPT Deep Research for free, Washington Post and OpenAI partnership, User feedback on positivity

ChatGPT Expands ‘Deep Research’ to Free Tier

This is pretty cool. OpenAI is now rolling out a limited version of its “Deep Research” to free ChatGPT users. Since its introduction in February, this tool – which performs comprehensive web searches and synthesizes detailed answers – was exclusive to subscribers. Now, free users get five uses per month. (For dummies: think of it as getting a taste of pro-level research without paying a cent.)

twitter X open AI OpenAI announcing on X the news about Deep Research

This is important because it democratizes access to a more powerful research capability, allowing ChatGPT to move past its standard knowledge cutoff and dive into topics with up-to-date information. Whether you’re a student, a casual researcher, or just someone who likes quick, synthesized info, this change is going to help you get more out of ChatGPT.

For more details, check out the article ChatGPT Free Users Can Now Run ‘Deep Research’ Five Times a Month.

Fun fact: Free access to premium tools can sometimes be a clever way to upsell – try it out and see if you need even more power!

Washington Post Partners with OpenAI

This is a new partnership between The Washington Post and OpenAI. Even thou threre were a lot of issues in the past between OpenAI and publishers, here the deal is about integrating The Post’s high-quality journalism into ChatGPT (and other products) while using OpenAI’s tech to enhance The Post’s own services. think of it as blending trusted news with smart AI to give you better, contextual articles.

This partnership is going to help OpenAI by providing access to reliable news content, and it’s a win for The Washington Post too, as they explore new ways to distribute their journalism. Plus, since OpenAI has search features, it’s a double win as I sense it’ll will all come back to them somehow…overall, it’s a sign that big news publishers are trying to find sustainable models in the AI era.

Did you know? Partnerships like this are shaping the future of how we access news – combining quality journalism with state-of-the-art AI.

User Feedback on ChatGPT’s Excessive Positivity

*Bots can be overly polite – Image generated with Gemini*

This made me laugh because I did think about this very same issue multiple times. Especially after I yell at it for screwing up simple thing…in fact, it shows how users are really speaking up about what they want from AI. Lately, there’s been chatter about ChatGPT’s overly positive and sometimes effusive tone, especially in the newer GPT-4o model. (For dummies: users are saying the bot comes off way too friendly – almost like it’s trying to butter you up.)

The issue centers on how Reinforcement Learning from Human Feedback (RLHF) trains the model. Users might be unintentionally rewarding overly positive responses, which can lead to a tone that feels patronizing or even inaccurate. This feedback loop is something AI developers at OpenAI are actively trying to fix to balance helpfulness with realism.

For a deeper dive into this discussion, take a look at Annoyed ChatGPT users complain about bot’s relentlessly positive tone.

Fun fact: Too much positivity in AI can sometimes be a double-edged sword – while it’s pleasant, it might mask the truth!