Generative Pre-trained Transformer

or better known as GPTchat. What’s it good for?

The technology is based on an artificial intelligence model called Generative Pre-trained Transformer 3 (GPT-3), which uses machine learning to generate human-like responses to natural language queries. Chat GPT3 is able to interpret user queries and respond to them without relying on rules or programmed scripts.

My blog has gone down the AI rabbit hole since the sudden interest invoked by this Chat Bot. Well this is only the 4th in a series that started with an arguement that AI is not very well suited to marketing.

But I find this the most interesting rabbit hole out there, so there will be much more forthcoming. But first; I was right!

I really started this post to pointing out that I was exactly right when I wrote “Im betting the short-term solution is that Microsoft will seemingly bungle everything and ChatGPT is not allowed to come into use in any way that would disrupt the (search engine advertising and surveillance) game. Not exactly strangling ChatGPT in it’s crib, but neutering or controlling it.”

So the narrative will now be “its racist” so that normies can be conditioned to beleive that AI isn’t ready ( which is new-speak for “isn’t going to be used just yet” ), but in fact the flood-gate has been opened and those who “got it” are now licensing the technology for their own purposes.

Griffin views ChatGPT as a way to streamline tasks for employees, enabling them to be more productive. “It will take an enormous amount of work that’s done today by people, and do it in a distinctly different, highly automated, efficient way,” he said.

What would those purposes be, exactly?

Well for one, any business based on rules will be fertile ground to disruptive with GTP3 AI. There will be the early adopters like Ken Griffin, CEO of Citidel (financial), and then there will be those who bought the narrative and will be caught flat-footed as they become (suddenly to them) obsolete.

Business That Exists on interpreting or following Rules

I have in mind CPAs and lawyers. Prime businesses to be obliterated. Their business is reliant on rules (laws are rules). Their business is advising how to follow rules or work to follow those rules.

Humans can’t handle the tax code of the US IRS as it exists. No one would argue against that statement. But a AI can. It just has to be fed the documents to digest, and presto, you would have the world’s best tax advisor/auditor money can buy. I gotta believe the CEO of Intuit has done just as Ken Griffin; license it and put every other service (and individual) out of business. First-mover advantage.

Consider this..

In 2012 The National Taxpayer Advocate came up with roughly 4 million words. Again at roughly 450 words per page, that comes out to around 9,000 pages. The National Taxpayer Advocate also noted that the tax code changed 4,680 times from 2001 to 2012, an average of once per day.

There are 654,375 actively licensed CPA in the US. That’s a whole lot of unemployed people.

How about Lawyers? Same, same, different.

Following this rabbit hole, would it not be better to put an AI to the task of auditing the tax code itself? These rules are written ad-hoc by people who don’t even have a grasp of the entirety of the tax code, so of course rules are convoluted and therefore they often conflict. That could be straightened out by an AI in short order and then you won’t need that expensive tax attorney.

For an insiders essay on “LegalTech” as its called, here is; GPT-3 and ChatGPT: Disrupting LegalTech. There is an interesting train of thought that emerges when reading such articles; accuracy & authenticity seem to be the dominant themes. I think that is only a nuance. The real issue is disruption, but insiders would rather not notice the elephant – of course. Lets talk instead about how it helps the “outsiders” like me. – that’s more like it. How accurate does it need to be?

For instance, if I could license OpenAI’s GPT3, I would immediately set out on the path of finding those rules from which I could profit by finding loopholes – we know they exist everywhere. They are created and cleverly hidden from easy public access to benefit friend/family/special interests. The only problem is finding them. A GPT3 AI could unearth those with easy. Or I could set out to find all the ways to qualify for entitlement programs (free money from the government) if I only had a GPT3 AI that I could feed all the massive, convoluted government regulations to.

How about Politicians? Same, same, different.

The rabbit hole leads to questions about government regulation itself. How about a GPT3 AI that could identify and block laws from passing that conflict with the constitution? Why would we need a Supreme Court? (who hardly follows the constitution anyway) Seeing how the government is not fit to regulate itself, a GTP3 AI could keep our politicians in line (in real time) better than voters can. There is a game-show in there somewhere.

There is just too much to disrupt but I find the “scent” to follow for disruption are “rules” and “laws” – areas that seem highly disreputable by GTP3 AI.