What You May Have Missed #13

ChatGPT and a whole world of implications / On the future of AI / Bad uses of language models / Meanwhile on the visual generative space

AI imagines Spain as a villain. Credit: Loubfreitas (TikTok)

ChatGPT and a whole world of implications

OpenAI’s ChatGPT vs AnthropicAI’s Claude

AnthropicAI has trained a ChatGPT-like model named Claude that apparently is better than ChatGPT. ScaleAI's staff prompt engineer, Riley Goodside, is our most outspoken source for now (I've looked for info but it seems there isn’t much online).

Riley Goodside @goodside

A few weeks ago I got early access to a new model from @AnthropicAI. Its name is Claude. To demo, Claude will give the first scene of a Seinfeld episode where Jerry receives the supposed key to AI alignment from @sama, ignoring dire warnings from Kramer. The prompt:

Here’s a side-by-side comparison between Claude and ChatGPT. An interesting reflection by Goodside: “On a closer read, the first bullet point in Claude's answer does not make sense. Proofreading these things is difficult, and it's getting harder” [emphasis mine].

Riley Goodside @goodside

Side-by-side comparison: @OpenAI's ChatGPT vs. @AnthropicAI's Claude Each model is asked to compare itself to the machine from Stanisław Lem's "The Cyberiad" (1965) that can create any object whose name begins with "n":

Here are some limitations. Kudos to Goodside for making a conscious effort to unbias his analysis.

Riley Goodside @goodside

As a counter-balance to all the positivity I'm posting about Claude (a new model from @AnthropicAI), starting a thread on limitations. First, for many calculation problems, hallucinations are just as bad as in ChatGPT — here's ChatGPT vs. Claude vs. a calculator: