Well, this shores up one of ChatGPT's biggest weaknesses, especially GPT4's. I wonder when we'll see the pace of refinement for this drop off.
If wonder if it will drop off.
ChatGPT can now call Wolfram Alpha
181–190 of 329 posts
Re: ChatGPT can now call Wolfram Alpha
#182Earlier quoted context omitted.
In terms of product innovation, OpenAI is now like Google and Google is now like Xerox.
The analogy is even more apt when you think that most of the things Apple and Microsoft became big with in the 90's - like mouses and GUIs - were invented as Xerox but they failed to monetize, just like the basis for GTP came from Google and they failed to monetize (externally). I don't think Google is as Xerox- or IBM-ified yet though, they might still bounce back with AI products.
Please anyone willing to answer as if you might be speaking to a golden retriever:
How important is a "lead" in this space?
I have been comparing ChatGPT and Bard a lot the last two days because this is all very fascinating, and from my datapoint-of-one perspective, it feels like ChatGPT is way, way ahead.
Then I start to wonder is this because ChatGPT captivated our attention so much the last few months that they've been able to inherently improve the product faster, and thus the flywheel starts spinning rapidly giving them even more of an advantage as more people use the service, giving it more to train off of?
I.e., does Bard (and any forthcoming competitors) fall way behind because they lost the slightest head start that then rapidly spirals into a competitive advantage?
Re: ChatGPT can now call Wolfram Alpha
#183I bet the Zapier integration is going to be an alignment/security dumpster-fire. Now, when a bot gets prompt-hijacked into a malicious personality, it can make API calls to other systems! Great news. Also, plug-in APIs will presumably be able to prompt-hijack the GPT instance that is calling them. I would not be surprised if we see someone get swatted through GPT this year, I think it’s very likely to be within the n…
What’s swatted?
Re: ChatGPT can now call Wolfram Alpha
#184The interesting thing is that, now that the code is written, OpenAI can probably point GPT-4 at it and say "now do that for this list of other web services" including, presumably, itself when it needs to recursively work on a problem.
Yes, this is the singularity plus some minor human involvement.
gg
Re: ChatGPT can now call Wolfram Alpha
#185OpenAI and Microsoft have been killing it not only in product but in the go-to-market work they've been doing long before launch. Really superb work. With seemingly every launch, they have a slate of really compelling launch partners already using the new service, giving them a big press bump (from the articles those partners write) and social proof, and of course the product is better because it's already been teste…
Yes, it's surprising me how quickly they're moving ahead. It's almost as if this is a plan of staggered launches since some year back coming into fruition. I'm not sure how else they could get weekly major announcements like this.
I'm wondering if they are just leapfrogging everyone by using the LLMs to write code faster.
Re: ChatGPT can now call Wolfram Alpha
#186Earlier quoted context omitted.
And also whenever the next iteration bootstraps off the Wolfram outputs so it no longer has to call Wolfram & can better generalize/learn thanks to it...
Wolfram himself laid out why this is unlikely to happen, https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-... >Or put another way, there’s an ultimate tradeoff between capability and trainability: the more you want a system to make “true use” of its computational capabilities, the more it’s going to show computational irreducibility, and the less it’s going to be trainable. And the more it’s fundamentally…
Many of the examples are things that a feedforward model like GPT-4 has plenty of capacity for, could do (eg all the real-world facts), and each example helps induce the transition between memorization and generalization so it will do. In some cases like multiplying large numbers, it can't no matter how many examples it's given (with a realistic number, anyway) - but every query firms up the ability to swap out Wolfram Alpha for any other calculator...
Wolfram is playing with fire here (https://gwern.net/complement) given that the output is going to, one way or another, assist training future AI models.
Re: ChatGPT can now call Wolfram Alpha
#187OpenAI is like Google before 2003, dazzling the world with amazing products backed by astonishing technologies that went beyond wild dreams of millions of people. And most importantly, OpenAI kept releasing new features at blazing speed with a few hundred people, just like early days of Google and Microsoft. This reminds me of stark contrast of Google's GSuite and Microsoft Office. Microsoft released new office every…
I don’t remember Google being this interesting. Ever. It was a good service and very useful, but nothing like this. It was not astonishing. The tech was mildly interesting, but most of all very useful.
Did you use Dogpile, Ask Jeeves or Yahoo search and try to get useful information out of them? Webrings were still a thing where a website about birds or something would have a list of other sites that had similar indie - and that was how you browsed the web as the search engines weren't good. Until Google.
Re: ChatGPT can now call Wolfram Alpha
#188In the end I think LLM's will not be used on their own much, and instead more as glue for bringing together a whole bunch of other technology and allowing it to be coordinated with plain English.
You can say the same thing about humans... our greatness over the rest of the animal kingdom comes from tool use. Microsoft researches just put out a paper about 'proto-AGI' in the last 24 hours or so, and tool use was a big part of the paper.
"Sparks of Artificial General Intelligence: Early experiments with GPT-4" -- https://arxiv.org/abs/2303.12712
Re: ChatGPT can now call Wolfram Alpha
#189I bet the Zapier integration is going to be an alignment/security dumpster-fire. Now, when a bot gets prompt-hijacked into a malicious personality, it can make API calls to other systems! Great news. Also, plug-in APIs will presumably be able to prompt-hijack the GPT instance that is calling them. I would not be surprised if we see someone get swatted through GPT this year, I think it’s very likely to be within the n…
What’s swatted?
Re: ChatGPT can now call Wolfram Alpha
#190OpenAI is like Google before 2003, dazzling the world with amazing products backed by astonishing technologies that went beyond wild dreams of millions of people. And most importantly, OpenAI kept releasing new features at blazing speed with a few hundred people, just like early days of Google and Microsoft. This reminds me of stark contrast of Google's GSuite and Microsoft Office. Microsoft released new office every…
This is pretty hyperbolic. Great products that were much better than what they replaced or competed with, but "beyond wild dreams of millions of people"? You'd think they cured cancer in 2003 or something...
Google Maps was pretty astonishing when it appeared.