@mikeknoop I think the analogy to compilers is a category error. Compilers are, for the most part, deterministic functions. Induction is more straightforward.
I think a far better analogy for LLMs is outsourcing via offshoring. I lived through that, and it had… mixed results.
I know people have worked really hard on making this all work about as well as it can in a sorta backward compatible way. But I can't imagine it being much worse if the goal was to make it as complicated as possible.
Beyond excited that @OpenAI ChatGPT is launching plugins today -- and to share @zapier is a launch partner! Via Zapier, ChatGPT can now interact with over 5,000 apps like Google Sheets, MS Teams, Salesforce, Gmail, and thousands more https://t.co/fizkD4rfMf
Excited to show off what I've been working days/nights/weekends on the last several months!
✨New: Zapier Natural Language Actions API
https://t.co/Jk3e0u0QVT
And "should" has no power on its own. Willpower is only good for an initial push but quickly runs out. If you're not making the "should" easier and the "should not" harder, then it's unlikely than anything will actually change.
@AmazonHelp Purchased from Amazon, delivered by Amazon. Last 4 digits of recent order: 8209. Last 4 of tracking ID: 4076. I’m signing off till tomorrow, er, later today. Thanks though. 👋
I’ve mostly had good experience with @amazon orders over the years, but apparently if they mess up sending you an item over $1000, you may find them holding your money hostage.
I’ve mostly had good experience with @amazon orders over the years, but apparently if they mess up sending you an item over $1000, you may find them holding your money hostage.