Linked List: June 4, 2023

Sunday, 4 June 2023

Introducing the Wavelength AI Bot Designer 

Richard Henry, writing for the Wavelength blog:

AI is a huge part of Wavelength today — 1/3 of all messages sent invoke the @AI bot in a conversation.

Today we’re launching an AI bot designer built into Wavelength. This lets you easily create custom AI bots — amazing lifelike personalities that you can interact with in your group chats or via direct message. You can choose from GPT-3.5 or Claude Instant v1.1 for the model that powers your bot. We’ve found that sometimes a particular model is much better for a character, so it’s worth trying both and comparing results. [...]

Make a fun bot? You can share it with others using a link, or make it public so that anyone can discover it. Make a fun character, a text-based adventure game, or anything else!

I’ve been testing this feature for a few weeks (in my previously disclosed role as an advisor to Wavelength), and it’s a lot of fun. For characters, I’ve found the Claude model, from Anthropic, to be better. Or at least more “in character”. Two fun bots I’ve made: Don Rickles and Triumph the Insult Comic Dog.

Here’s a screenshot showing the difference — same prompt/bot definition, just switching from GPT 3.5 to Claude. It’s fun to get trivia question answers with a bit of personality.

Ted Chiang on AI: ‘The Machines We Have Now Are Not Conscious’ 

Science fiction writer Ted Chiang, in an interview with Madhumita Murgia for The Financial Times (Archive.is link):

Chiang’s main objection, a writerly one, is with the words we choose to describe all this. Anthropomorphic language such as “learn”, “understand”, “know” and personal pronouns such as “I” that AI engineers and journalists project on to chatbots such as ChatGPT create an illusion. This hasty shorthand pushes all of us, he says — even those intimately familiar with how these systems work — towards seeing sparks of sentience in AI tools, where there are none.

“There was an exchange on Twitter a while back where someone said, ‘What is artificial intelligence?’ And someone else said, ‘A poor choice of words in 1954’,” he says. “And, you know, they’re right. I think that if we had chosen a different phrase for it, back in the ’50s, we might have avoided a lot of the confusion that we’re having now.”

So if he had to invent a term, what would it be? His answer is instant: applied statistics.

My puerile mind is tempted to make a joke that tacking on “system” would make for a fun acronym, but I shan’t crack that joke, as I think Chiang makes a strong point here. What we have with these LLMs isn’t low-level intelligence but rather high-level applied statistics that creates the powerful illusion of low-level intelligence.

See also: Chiang’s very short story “What’s Expected of Us”, referenced in the interview.