Artificial Intelligence: Chat GPT, BARD, etc

[SemiSalt]

Google has added BARD, their AI engine, to the mix of Great New Ideas. It comes with the specific warning that it may give incorrect answers. So I asked Bard a question: What is a good prismatic coefficient for a planning dinghy?

And Bard answered:

A good prismatic coefficient for a planning dinghy is between 0.2 and 0.25. This means that the boat is relatively long and narrow, with a high ratio of length to beam. This shape helps the boat to plane, or lift out of the water, when it is moving quickly.

A higher prismatic coefficient will make the boat faster, but it will also make it less stable. A lower prismatic coefficient will make the boat more stable, but it will also make it slower.

The best prismatic coefficient for a particular boat will depend on a number of factors, including the weight of the boat, the power of the engine, and the type of sailing that the boat is intended for.

I was somewhat impressed that it was trained with so much material that it could make a coherent sentence about prismatic coefficient. I suppose there is a design book or two caught up in Google Books. OTOH, if anyone has a lines plan with a PC of 0.2, I'd like to see it.

[Slick470]

Semi, that's pretty cool, and a bit scary all at the same time. As a kid who grew up with the Terminator and then the Matrix movies, I'm pretty skeptical about our future machine overlords.

[Steele]

I hope the benevolent future of these things greatly surpases the potential evil. I recently listened to an NPR report of GPT being used to help blind people. Not only can it read things like a menu using a smart phone, but the individual can ask questions like "I'm in the mood for a chicken dish,what are my options?"

[BeauV]

I'm triggered by this thread!

By that, I mean that as a young programmer, I went full steam into helping to build all sorts of online communities from The Well through much more modern versions of the idea. My naive views about the "goodness of most people" were destroyed within two years. This was due to all sorts of issues, but I believe the most important one is that they evolved to allow anonymous users. (Long arguments on both sides of this debate online and worth reading about. Google "Larry Lessig, Harvard Law School for the opposition view to my own.)

The primary reason for ever allowing anonymous posting online is to protect those who wish to blow the whistle on the people in power. Obviously, retaliation by the "people in power" is quite real around the world, including in the US. (Retaliation by individual thugs is also a real problem.)

So long as every use of Chat GPT is required to identify itself as a piece of software, I'm fine with it recommending menu selections and guiding blind people.

I am not fine with using this technology to create millions of bots that persuade the feeble-minded or naive that they are talking to a real person. Of course, the genie is already out of the bottle. I suppose it is just one more reason to continue to educate people that the cannot TRUST anything they read anywhere even if it now talks back to them in a way that makes them feel it's human. The political consequences of extending "free speech" rights to this kind of software are some of the worst things that technology has ever done. This is de facto what we've done with Chat GPT. Ugh! I need a drink.

[TheOffice]

I tried Bard on 2 very different questions.

First I asked it when was the best tie to sail from Bermuda to the Azores. I got a fairly impressive answer with lots of disclaimers.

Then, I asked it a legal question I had just researched. I got a simplistic and incomplete answer without any citation to law.

Its a work in progress.

Joel