Welcome to the new community!
To get things started, I’m curious what sorts of things everyone is using the current generation of AI for.
What tools do you find the most useful today, and how are you putting them to use?
I would love to hear how people are adapting AI to be helpful in a personal or unexpected way. Whether you use it for work, play, or something completely unique, how are you making it work for you?
- Which tools do you consider the “best” right now?
- What sets them apart from other tools you have tried?
- Do you stick to a single assistant, or do you use a combination of different tools?
Looking forward to reading about your experiences.
Just to put it out there, I’m under no delusions about the state of AI. I know it isn’t perfect (let’s be honest, far from it), but I think it’s a tool that can be useful in its current state and I enjoy experimenting with it.
Aside from making dumb memes for group chats, I’ve been using it almost like a frontend for search engines. Now that some models can query search APIs directly, it’s really helpful to bypass the minefield of modern sites. Even with browser extensions the open web can be annoying, but AI is often able to present the information I’m looking for more cleanly.
Sports make a good example. For something I follow casually like the NFL, I can just ask what teams are still left in the playoffs? and get a clean TLDR without chasing the dangling carrot from a search engine.
People talk about AI inaccuracies, which is a valid concern, but the traditional web has never been blindly trustworthy either. It’s full of clickbait, reactionary takes, and outright lies. For me, using AI to filter that noise is just a practical way to deal with it, like an advanced adblocker.
That said, it is important to remember that there is no such thing as a single source of truth. Whether it’s an AI model, a major news outlet, or Wikipedia, assuming any source is infallible is a mistake. The citations make it easier to spot inaccuracies in responses, but the responsibility to double-check information is still there.
It’s a tool. Searching pubmed only scans manuscript titles and abstracts, while GPT can scan into the manuscripts , excel tables, data files, supplemnents, etc. for very specific things like a specific reagent or experiment.
We use Alphafold for protein biochemistry modelling daily and virtual experimental designs.
But I would never use the summaries or trust a response to a question because it’s not accurate.
I find ChatGPT with internet access and deep “reasoning” to be very useful at asking questions; i.e., “Find me articles that talk about $thing”. Since it puts references, I can go to them and make sure it’s not halucinating. Basically, a glorified search engine, where it “underestands” more or less the question, instead of me having to search with keywords.
Removed by mod
