Claude 3 Opus went live on Monday as the latest company to enter the LLM (Large Language Model) space, and I had to give it a try. Of course, the first question I always ask is, “Who is Tom Nault, who now lives in Missouri?” Claude chickened out and didn’t want to give me much of an answer. I prompted it with a bit more personal information, and it still couldn’t tie any data together, so it’s nowhere near as resourceful as the other LLMs just yet. The others will try, even when they get so much wrong. I use my name because I’m the only person who can read a bio and know 100% what’s true and what isn’t.

Meanwhile, GPT-4 is fairly good but still has a long way to go. All of these are still in their infancy, and I use them every day as a proofreader without changing my voice. The second it changes my voice to its own, I stop the editing and start over. Quora’s Poe is probably the worst at being helpful. Its suggested images from AI renderings are ridiculously stupid. Here is an example.

Much like the disastrous Google Gemini launch, this image tried to appeal to left-hand and right-hand drive images, so it compromised by placing the steering wheel in the middle and then made it impossible to turn, with the flat bottom resting directly on the center console.

Notice the center armrest, where no arm could possibly go, and the right-side seat bolster and left-side seat lump for your shin bone. I have no idea what's going on next to the seat.

This is an attempt to weave "Wokeness" into a geographically neutral position. It's not a left or right hand; it's a compromise. It's a great image because it shows how they see the world. It's bad training. As the old saying goes, "garbage in, garbage out."

I've been screaming about the horrendous side of "woke" and DEI (Diversity, Equity, and Inclusion) and what it does to companies. I've also written extensively about what happens when you drive debate underground and the consequences. Pirate Wires did a great job of covering it. I also think Vivek Ramaswamy did a great job of summarizing.

Here are some of the images generated by Google's Gemini that has everyone laughing at them. I believe that’s a black George Washington.

I described myself to GPT-4 to see what it rendered.

This just wasn’t going to go in a direction that was closer to what I looked like. I finally had to just get blunt. It can’t render an image from a photo of me, so I gave it even more info.

Now we’re way off… I’ll give it one last try.

Looks just like me! Notice anything similar in all images? I never specify skin color so it makes the race ambiguous. All are male stereotypes. It’s very cool that AI can do this from just words, but there is a long way to go here.