GPT Plugins: Existential No-Code Threat?

API
1

Bold headline, right? An existential threat to what exactly? Well, um, everything compute-related. But let’s narrow this missive down to Airtable itself (for example).

Unless you live in a cave, you’re aware of ChatGPT. It’s in the process of roto-tilling pretty much everything compute-related. While not generally seen as a threat to much, we all see the writing on the wall; the computing world is changing in profound ways with LLMs. The rate of advancement AGI has demonstrated in recent weeks and months is obvious.

Yesterday, OpenAI opened the [next] kimono to reveal yet another massive game changer; GPT Plugins.

CleanShot 2023-03-24 at 00.40.39@2x
CleanShot 2023-03-24 at 00.40.39@2x1790×576 103 KB

Let that sink in.

While all aspects of this new GPT capability are significant, one flies in the stratosphere of all other possibilities - to run computations. While everyone is drawn toward “access up-to-date information” in a mesmerizing mental trance like a Star Wars tractor beam, the most stunning part of the announcement will be vastly overlooked.

We crave the day when GPT can see the Interwebs in real-time. GPT plugins make this happen in an instant with yesterday’s announcement. But as cool as that is, it will ultimately be seen as a distraction away from the tectonic plates that are about to collide.

Run Computations

What will the future be like when you can ask a system to build a process (as code), provide it with some data, and then run that process against it?

As demonstrated here, OpenAI has revealed:

  • ChatGPT is a full compute stack.
  • ChatGPT has CPUs and GPUs available for everyone and any reasonable objective.
  • ChatGPT has a file system.
  • ChatGPT can transform your words into processes and execute them.

Let that sink in a little deeper.

Imagine Zapier comes along and says to OpenAI -

Hey, with that new plugin architecture, what if we made it possible to use our five thousand connectors in natural conversations to get data, documents, and other information that the LLM can use to answer questions, write reports, and otherwise serve as the dynamic gateway to everything the LLM wasn’t trained on?

This hypothetical is not fiction; it’s a done deal.

Airtable Threat?

And then Zapier says -

BTW, what if we could also provide the database backend for all manner of natural languages to store or transform data?

This part is hypothetical but a likely reality in the near future. I warned over here not long ago that Zapier is coming for more than your adhesive dollars; it wants your data.

Suppose you can interface with any datastore, retrieve data, perform computations on that data, and build your apps’ computational aspects by using your natural prompts in any language; who (or what) stands to be disrupted?

For starters - everything related to integration. GPT can understand how to POST or GET data to and from any endpoint based purely on API documentation. It has this capacity today. GPT plugins take this innate integration ability to another level - they can execute data interchanges, build log files, track performance, and report exceptions.

Let that sink in.

1 Like
2

From my point of view, and what I find even more terrifying, is the threat that exists because there is no way to detect false negatives in a system where 100% of the compute is abstracted. Malicious actors, and subtle flaws rely on false negatives to exist.

1 Like
3

But isn’t all compute now abstracted for the most part? We live in a world where compute operations rarely occur localized with the inputs and outputs, and yet - our business and personal processes are not rife with false negatives. If that were the case, advanced medicine, space travel, agriculture, and manufacturing would be chaotic. That’s not what’s happening. Largely, humanity is thriving with the aid of computing systems.

Perhaps, but you’d have to give me an example to help me understand your point.

I struggled with the word “existential” in this topic and why it is predicated by “no-code”. I’m calling attention to the existing and thriving no/low-cold tech segment that has dominated the democratization of information movement for about a decade. Based on advancements in AI, it appears disruptive applications and services that have provided real value by abstracting form from function are, themselves, about to be disrupted by the ultimate abstraction - natural language.

1 Like
4

“But isn’t all compute now abstracted for the most part?” Not at all. The code we’re talking about is first class. Up until now if I wanted to connect to an external api I would write code for it. As far as false negatives, think about antivirus software that misses a bit of malware on your computer even though it has the signature for that malware. The antivirus software falsely reported a negative scan. Another example is in ossec where a malicious actor changes runtime configuration in memory but your ossec scanning software doesn’t report it because it only looks at files on disk.

5

I see. Interesting analogy. But, that’s a different issue. Fearing what AI might falsely miss and using AI to disrupt the no/low-code database segment are two very different topics. You should write about that fear because they are valid risks.

My post suggests that given a future where thousands of GPT Plugins exist in an app-store-like experience, some of these apps will show users (as an example) how they can build database apps to solve business problems without ever having a database platform.

6

Good read. Thank you.

I think it’s completely possible for someone without coding experience to “talk” a very simple interface to display the contents of a database. I think this is similar to using Wix, Wordpress, or any other sort of “no-code” platforms. Personally, I think it’s a bit of a hype word when it comes to LLMs, and it’s not utilizing them to the best of their ability. These features exist with the already mentioned companies.

A huge, and vital part of prompting is being knowledgeable enough to prompt correctly. Or being critical enough to evaluate the response, and make the determination if it’s useable. In terms of having an application which has a database, and perform calculations on it, we’re talking about an intricate process that seems very straightforward, but truly isn’t if done right. In fact, it would be the knowledgeable one that could create it as a straightforward process.

Even considering that, what benefits does talking do, when gestures and pre-defined actions work just as well? Templates exist and already do these things.

As is with Wix, or Wordpress, there is such a great reliance on plugins, or essentially code that other people have written. Not only do these plugins inherently cause limitations, they also cause a very bloated web page as they are not properly used or optimized.

Will it cause a huge influx in “no-code” applications? For sure. Reminds of me of when websites could be made instantly for free using Geocities. Just because it’s easier to do, I really don’t think it will make much of a difference.

1 Like
7

I don’t see GPT Plugins as anything more than a new interface to SaaS offerings. Just like Zapier is a web-based interface to SaaS. But because it’s and LLM, it might be easier than the web based interface. And because it’s an LLM prone to hallucinations, the result is going to be sloppier.

The real cool factor is the constant interaction with conversational style, as seen in the recent Wolfram demo.

Having access to data in a database? Yawn. Something needs to put all that data in there to be usable, and it’s not the user, it’s another SaaS provider.

Retrieval plugin? Yay! Finally I don’t have to explain embeddings anymore! But a competent dev already can do this, nothing new. But good in that it lowers the barrier-to-entry for embeddings, but the consumer will pay through the nose for all those surrounding SaaS providers doing it for them!

3 Likes