I downgraded to Team subscription, I think this is gonna make me upgrade to Pro again.
ChatGPT agent: bridging research and action
71–80 of 508 posts
Re: ChatGPT agent: bridging research and action
#72I do not know what an agent is and at this point I am too afraid to ask.
It's just a ~~reduce~~ loop, with an API call to an LLM in the middle, and a data-structure to save the conversation messages and append them in next iterations of the loop. If you wanna get fancy, you can add other API calls, or access to your filesystem. Nothing to go crazy about...
Technically it's `scan`, not `reduce`, since every intermediate output is there too. But it's also kind of a trampoline (tail-call re-write for languages that don't support true tail calls), or it will be soon, since these things loose the plot and need to start over.
Re: ChatGPT agent: bridging research and action
#73For me the most interesting example on this page is the sticker gif halfway down the page. Up until now, chatbots haven't really affected the real world for me†. This feels like one of the first moments where LLMs will start affecting the physical world. I type a prompt and something shows up at my doorstep. I wonder how much of the world economy will be driven by LLM-based orders in the next 10 years. † yes I'm awar…
It went viral more than a year ago, so maybe you've seen it. On the Ritual Industries instagram, Brian (the guy behind RI) posted a video where he gives voice instruction to his phone assistant, which put the text through chatgpt, which generated openscad code, which was fed to his bambu 3d printer, which successfully printed the object. Voice to Stuff. I don't have ig anymore so I can't post the link, but it's easy…
Re: ChatGPT agent: bridging research and action
#74It's very hard for me to imagine the current level of agents serving a useful purpose in my personal life. If I ask this to plan a date night with my wife this weekend, it needs to consult my calendar to pick the best night, pick a bar and restaurant we like (how would it know?), book a babysitter (can it learn who we use and text them on my behalf?), etc. This is a lot of stuff it has to get right, and it requires a…
This problem particularly interests me.
One of my favorite use cases for these tools is travel where I can get recommendations for what to do and see without SEO content.
This workflow is nice because you can ask specific questions about a destination (e.g., historical significance, benchmark against other places).
ChatGPT struggles with: - my current location - the current time - the weather - booking attractions and excursions (payments, scheduling, etc.)
There is probably friction here but I think it would be really cool for an agent to serve as a personalized (or group) travel agent.
Re: ChatGPT agent: bridging research and action
#75I think there will come a time when models will be good enough and SMALL enough to be localized that there will be some type of disintermediation from the big 3-4 models we have today.
Meanwhile, Siri can barely turn off my lights before bed.
Re: ChatGPT agent: bridging research and action
#76The "spreadsheet" example video is kind of funny: guy talks about how it normally takes him 4 to 8 hours to put together complicated, data-heavy reports. Now he fires off an agent request, goes to walk his dog, and comes back to a downloadable spreadsheet of dense data, which he pulls up and says "I think it got 98% of the information correct... I just needed to copy / paste a few things. If it can do 90 - 95% of the…
Of course, Pareto principle is at work here. In an adjacent field, self-driving, they are working on the last "20%" for almost a decade now. It feels kind of odd that almost no one is talking about self-driving now, compared to how hot of a topic it used to be, with a lot of deep, moral, almost philosophical discussions.
Re: ChatGPT agent: bridging research and action
#77I'm not so optimistic as someone that works on agents for businesses and creating tools for it. The leap from low 90s to 99% is classic last mile problem for LLM agents. The more generic and spread an agent is (can-do-it-all) the more likely it will fail and disappoint. Can't help but feel many are optimizing happy paths in their demos and hiding the true reality. Doesn't mean there isn't a place for agents but rathe…
>> many are optimizing happy paths in their demos and hiding the true reality
Yep. This is literally what every AI company does nowadays.
Re: ChatGPT agent: bridging research and action
#78One the one hand this is super cool and maybe very beneficial, something I definitely want to try out. On the other, LLMs always make mistakes, and when it's this deeply integrated into other system I wonder how severe these mistakes will be, since they are bound to happen.
This. Recently I uploaded screenshot of movie show timing at a specific theatre and asked ChatGPT to find the optimal time for me to watch the movie based on my schedule. It did confidently find the perfect time and even accounted for the factors such as movies in theatre start 20 mins late due to trailers and ads being shown before movie starts. The only problem: it grabbed the times from the screenshot totally inco…
Honestly might be more indicative of how far behind vision is than anything.
Despite the fact that CV was the first real deep learning breakthrough VLMs have been really disappointing. I'm guessing it's in part due to basic interleaved web text+image next token prediction being a weak signal to develop good image reasoning.
Re: ChatGPT agent: bridging research and action
#79Any idea when we'll get a new protocol to replace HTTP/HTML for agents to use? An MCP for the web...
Re: ChatGPT agent: bridging research and action
#80The "spreadsheet" example video is kind of funny: guy talks about how it normally takes him 4 to 8 hours to put together complicated, data-heavy reports. Now he fires off an agent request, goes to walk his dog, and comes back to a downloadable spreadsheet of dense data, which he pulls up and says "I think it got 98% of the information correct... I just needed to copy / paste a few things. If it can do 90 - 95% of the…
I think this is my favorite part of the LLM hype train: the butterfly effect of dependence on an undependable stochastic system propagates errors up the chain until the whole system is worthless. "I think it got 98% of the information correct..." how do you know how much is correct without doing the whole thing properly yourself? The two options are: - Do the whole thing yourself to validate - Skim 40% of it, 'seems…
This depends on the type of work being done. Sometimes the cost of verification is much lower than the cost of doing the work, sometimes it's about the same, and sometimes it's much more. Here's some recent discussion [0]
[0] https://www.jasonwei.net/blog/asymmetry-of-verification-and-...