Fascinating Assessment on the Current State of AI

[JohnL]

Back to the original topic, I am beginning to suspect AI is in some sense impossible, that it will never be truly intelligent, just better at mimicking human behavior. Is it possible the "wow" factor will wear off and we will adjust our valuation of the type of work AI is capable of producing downward to account for the built in falsity. When the printing press was new it had the power to overthrow governments but eventually people developed a more analytical approach and learned not to trust everything in print. Now the garbage is full of print barely read or not read at all. Maybe the net effect on AI will be to value or highlight the things that are real.

There will always be place in the world for genuineness, humanness, and personal touch. Some people may get along without this for a period of time but I expect their worlds to eventually dry up or come up short.

I hope your prediction about an AI future is correct! Until we get there, we as a society may have some really dysfunctional decades.

I hope you're both correct and the novelty fades after only a few years instead of decades.
This story was really upsetting. Who's doing the programming? Who vets the programmers. Who watches the Watchers who vet the programmers.
‘I flipped out’: Watch mom toss Alexa after A.I. demands 4-year-old daughter answer super creepy question

[Bootstrap]

Back to the original topic, I am beginning to suspect AI is in some sense impossible, that it will never be truly intelligent, just better at mimicking human behavior. Is it possible the "wow" factor will wear off and we will adjust our valuation of the type of work AI is capable of producing downward to account for the built in falsity. When the printing press was new it had the power to overthrow governments but eventually people developed a more analytical approach and learned not to trust everything in print. Now the garbage is full of print barely read or not read at all. Maybe the net effect on AI will be to value or highlight the things that are real.

It's not all or none.

Some people definitely realize this already. In my own work, I have pipelines that call both AI and traditional Python code. Some of the Python code is also generated with AI.

If I generate code using AI, I generate twice as much code for automated tests to see if the AI-generated code works, and run those automated tests whenever something changes. This is called Test Driven Development (TDD), and it's traditional, but writing the tests is now much, much easier.

If I am creating resources with AI, I use databases with trusted information (with MCP), and I save the result of each step so that it can be inspected by both humans and AI.

If I research something and AI is part of the picture, it is never the whole picture. If I write something, I often write it myself or write a list of bullet points, then take what AI gives me and rewrite it, perhaps choosing only parts of what AI gives me (it's quite wordy and redundant sometimes), then ask AI to criticize what I wrote.

If you cannot tell if the result is good, asking AI to create something is a recipe for disaster. Every day at work, AI gets things wrong many times. So do I. Often, AI catches mistakes I make. I also catch mistakes AI makes. And the interaction is useful - defending and refining what I think, being forced to go deeper, etc. is part of the value.

[Bootstrap]

I want to command the computer like Captain Kirk in Star Trek. I don’t want to serve the computer like Dave in 2001: A Space Odyssey.

What is good and true and important does not depend on the tools we use to produce it. We can evaluate the result. When we do, we should carefully consider what it does for us as humans.

Mike needs ads to bring in customers. Using AI is cheaper than a human designer, and takes less of Mike's time and energy. The goal is still the same: to bring customers into the store and sell food. Mike still needs to evaluate the ad and see if it seems good, then see if it brings people in.

If AI is used to produce materials for group Bible study, the goal is the same: we want people to be engaged in understanding and experiencing the text, accurately, without imposing other things onto the text, to draw them closer to each other and to God, to help people learn to apply Scripture to their own lives. The same people producing such materials now may be best positioned to evaluate the ones AI helps produce. Group leaders can see how they affect their groups in practice.

[mike]

I have avoided “Internet marketing” as a business since there is really no way to do so effectively without engaging in dishonesty, and I’d offer a worse service than my competition which is willing to engage in dishonest behaviour.

John, I've PM'ed you a link to an ad with a ChatGPT-generated graphic from my business. I'm interested to see if you think this ad contains something dishonest. If anyone else would like to evaluate the ad, PM me.

JohnH, are you ghosting me, paging JohnH