I signed up for the Google Bard beta and ChatGPT to compare them. These are two of the new Large Language Models - chatbots - that have sparked a lot of interest and concern lately in artificial intelligence. So far I have just used them to generate jokes to put in my kids’ lunchboxes.

On a desktop browser Google Bard not only tells you a joke; it explains why the joke is funny.

from mastodon:

% tell me a joke. Sure, here is a joke: What do you call a fish with no eyes? Fsh! This joke is funny because it is a play on words. The word "eye" sounds like the letter "I", so when you take away the "I" from the word "fish", you get the word "fsh". This is a silly and unexpected answer, which makes it funny.

aaronjustaaron on mastodon replies 'It's funny because this is what's going to kill us

One thing to know about these Large Language Models is that they don’t answer a question that you ask them. They answer “What would an answer to this look like?”

This can be very misleading as it may misquote information or make up sources. It’s doing its job; quotes and sources are part of a good answer to a question, but this isn’t something we expect when we ask a question- “Here is an answer backed up with made up statistics and quotes.”

I’m not sure what problem these chatbots are trying to address. As Barack Obama has said in a 2020 interview in the Atlantic “If we do not have the capacity to distinguish what’s true from what’s false, then by definition the marketplace of ideas doesn’t work. And by definition our democracy doesn’t work. We are entering into an epistemological crisis.”

We are in the crisis. How do we know things? If we look at pizzagate, the rise of Qanon, or the Indian WhatsApp lynchings- people have lost the ability to distinguish true and false. People equate facts from reputable sources with viral fictions on social media. Cite wikipedia or NPR to a conspiracy theorist, and their response is to attack these sources (as ‘fake news’ or ‘big tech’) and redirect to someone’s blog or podcast to learn the truth. How can communication be possible? How can our democracy work?

So far this is artificial intelligence generating words. What happens when it starts generating sound, images and video? Metahuman can today take a single image and turn it into a realistic character. Tools already exist that can turn a five minute sample of speech into a generative speech engine. RollingStone is already reporting of generative porn hitting reddit, and of porn being generated from images without the subject’s consent.

We have already seen the rise of spoofing attacks where an attacker pretends to be the CEO of a company and instructs an employee to wire money. We have seen the spoofing attack where the attacker pretends to be a friend or family member stuck somewhere and needing money. Fake videos, text and speech can make all these attacks more deceptive.

Recent water cooler talk at the office was split. In an extremely non-scientific poll of two, half were optimistic, saying “I can make a bot and send it to meetings!” and the other half concerned- “I don’t want to see what they do with this in middle school.”

By the way the chatbots are fairly useless at creating jokes. They just take them from /r/Dadjokes. Pressed, they will bail with “I apologize for not being able to generate an original joke that you haven’t seen before. I am still under development and learning how to be funny. I will keep trying and hopefully I will be able to make you laugh soon.”

Hopefully.