Toolformer: Language Models Can Teach Themselves to Use Tools

Here is a little GIF demo: <a href="https:&#x2F;&#x2F;twitter.com&#x2F;timo_schick&#x2F;status&#x2F;1624058382142345216" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;timo_schick&#x2F;status&#x2F;1624058382142345216</a><p>The possibilities of this line of research are endless. When language models can call APIs and&#x2F;or use UIs, they can become the unified natural language interface to any software, any website, any app. See also <a href="https:&#x2F;&#x2F;www.adept.ai&#x2F;act" rel="nofollow">https:&#x2F;&#x2F;www.adept.ai&#x2F;act</a>.<p>Siri and Alexa and Google Assistant are dead ends. Language models trained to use tools will finally be able to start delivering on the promise of a software assistant that works. And eventually, they will be a key part of robots that accept natural language commands to perform everyday tasks in the real world, like this: <a href="https:&#x2F;&#x2F;sites.research.google&#x2F;palm-saycan" rel="nofollow">https:&#x2F;&#x2F;sites.research.google&#x2F;palm-saycan</a>

Just like how humans progressed technologically.<p>Each individual human has information that is considered specialized (a narrow model). Without communication there is no way to access these specializations. And written and spoken language is just the best way to communicate we&#x27;ve come up with (so far).<p>Feels like these language models will be the glue that hold all the narrow models together, and can build on top of to create new narrow models.

This is the kind of brilliant idea that seems obvious in hindsight. The API call becomes just another kind of text for the LM to predict.<p>The most impressive part (to me) is that the LM was able to generate its own training data starting from &quot;nothing more than a handful of demonstrations for each API&quot;. That sounds like a technique worth learning.

I&#x27;ve gotten chatGPT powered chatbot to call a wikipedia API I set up for it. It needed to output &quot;!wiki {query}&quot; which would result in it getting the results as its next prompt, which it could then summarize to the user.<p>I was really impressed by how easy it is to get it to properly use such a thing, or the commands of the chat platform I was using.

I&#x27;m not in the ML field, but for those in the know: is this a step in the direction to enable LLMs to train themselves on new data? If not, how far are we from that stage?<p>The shortcomings of current versions is that they&#x27;re trained on old data, and that training takes a very long time. Having them train in the background and continually update their capability would be a major breakthrough. Or unleash Skynet, but cool nonetheless. :)

This is what Wolfram asked for, but he wanted a monopoly on the &quot;tools&quot; <a href="https:&#x2F;&#x2F;writings.stephenwolfram.com&#x2F;2023&#x2F;01&#x2F;wolframalpha-as-the-way-to-bring-computational-knowledge-superpowers-to-chatgpt&#x2F;" rel="nofollow">https:&#x2F;&#x2F;writings.stephenwolfram.com&#x2F;2023&#x2F;01&#x2F;wolframalpha-as-...</a>

I was waiting for a proof of concept like this! IMHO the next-wave of GPT-3 productization will involve mapping problem and solution domains to a text-based intermediary format so that GPT-3&#x27;s generalization abilities can be applied to these problems.

Have they try modeling around reward system? The basis of intelligence of living things are what&#x27;s required in order to survive. There has to be some ways to punish the models with some kind of death equivalent.

For people wanting to play with that: this is very close to langchain&#x27;s agents system (their documentation has a very impressive demo using both a calculator and Google searches as tools available to a language model).

If you are shocked about how much trust the tic toc generation yields to data horders, wait until you see them giving AI full access to their computer.

The next step is to get the model to evaluate its own failure modes, decide on a fix, write the code for it itself (calculator, datetime operations, external APIs that it knows of, or any other code for returning or transforming text) and then have the tool evaluate and learn to use the tools it&#x27;s written for itself.<p>A language model that could successfully bootstrap itself into using a web browser would be a dangerous thing.

Fantastic!<p>FWIW, ChatGPT as-is is good enough to know which (of a given set) of &quot;tools&quot; to use. I&#x27;ve had great fun doing prompt engineering: first asking it to pick which of a set of functions might be necessary to solve a problem, second prepending the list of selected functions and asking it to generate code.

I really like that they use GPT-J for this - significantly less weights than GPT and open source!

Is it possible to teach GPT3 to do something like this through prompts? Like, responding with an API call to get information when it doesn&#x27;t have enough?<p>The closest I&#x27;ve gotten is having the model output both the be API call and a hallucinated response

And i guess the bots will eventually need to search Google and browse around Github issues, and eventually put some comments there, just like a normal human.<p>Knowing some tools don&#x27;t work out of the box is some kind of high intelligence.

Very cool project. I guess LangChain has some code related to this, but I don&#x27;t think this team published their code anywhere

And there goes my evening . . .<p>Thanks for sharing!

All I want is an AI that lives in the browser and automatically does timezone conversions into my local time.