Today’s issue of More Than Meets the AI:
- 3 wishes for generative AI (676 words)
- A tribute to humanity (188 words)
- Love and War (39 words)
Whenever people talk about the future of AI, it’s always “the singularity this” and “artificial general intelligence that.” But I’m more interested in what’s happening right now in what I call the “pre-AGI” era.
These are the fun days when there’s nothing ahead but potential. Sure, that potential includes the possibility for harm. But it also includes the possibility for positive, life-altering change at the global scale.
Today, I want to talk about some of the quality of life upgrades that I believe LLMs could offer us in the near-term future (using currently available technology) without having to become full-on AGI.
Wish #1: A bouncer for public video game spaces.
Problem: I like playing games online, I don’t like dealing with jackasses.
Status Quo: If someone I’m in a game space with is violating the rules, I have to stop what I’m doing to deal with the problem by blocking, reporting, kick-voting, or leaving. The only way to avoid jackasses in public gaming spaces is to avoid public gaming spaces.
AI solution: Real-time audio-to-text language processing run through an LLM as a query, for an unspecified number of minutes after a user enters a gaming space. If the user’s interactions with the system trigger the LLM’s guardrails, the user is placed in a social quarantine.
In order to avoid offenders creating dummy accounts to get around bans, status quo content analysis can determine if a user’s voice or text expressions match those of previously banned accounts.
Best part: Those who are adhering to the TOS and community guidelines they agreed to upon creating an account don’t have to address troublemakers themselves.
Wish #2: An environmentally-aware trip companion.
Problem: I have to look up information on my surroundings when I travel because I can’t afford to hire a local expert to hang out with me 24/7 when I leave home.
Status Quo: Buy a travel guide, do an internet search, ask a local.
AI solution: Something like an “always-on” Google Maps assistant that periodically generates interesting commentary about your surroundings. For example, if you went for a walk in your neighborhood it might chime in with “the house on the corner was just listed for sale this morning” or “it looks like it will start raining in 20 minutes, would you like to head back home now?”
And if you went on vacation in Paris it might call out historical landmarks, popular restaurants with current availability, or even draw attention to travel concerns such as road construction or attraction closings.
Best part: Once developers figured out how to ping for interesting/useful data, it would be a matter of relative ease to fine-tune an LLM and interface to present the data to travelers in such a way as to make it feel like you’re on the road with a trusted, educated companion.
Wish #3: A model that parses global media in order to hallucinate personalized multimedia networks.
Status quo: Streaming music through music services, TV/movies through TV/movie services, and games through gaming services in separate, detached experiences requiring individual logins, passwords, and interfaces.
AI solution: A generator that creates a bespoke interface for experiencing media, complete with optional scheduling (so you can enjoy all your favorite media spread out across different days and time slots just like old-school cable television).
This wouldn’t create new content, it would just take the content you have access to and make it a bit more fun for those times when you’re not sure whether you want to listen to music, play a game, or watch something.
Best part: The whole point of such an interface would be to remove the intellectual burden that comes with opening and navigating half a dozen different interfaces. You’d open a single app and all your current legal media options would be hallucinated into a convenient “TV guide” style interactive interface.
Savvy readers will notice that each of these wishes is for a very narrow system trained for a specific purpose. That’s where I see the most value for LLMs in the near-term.
A tribute to humanity
AI is really good at a lot of stuff. But, if you want to get to the heart of its flaws, just ask an LLM to generate lyrics in the style of your favorite artist or band. To test this assertion, I spent about 30 minutes today asking ChatGPT to give me lyrics for various things in the style of the Dead Kennedys.
Don’t worry, I’m not going to share them here. Unsurprisingly, they were all terribly mediocre. None of them were bad enough to be funny and none of them were good enough to be remarkable.
I’m convinced that we’ll reach AGI itself before we invent a machine that can top the lyrics to “Dog Bite” by the Dead Kennedys.
That may say more about me than it does about AI, but the reality for me is that the fact they were written by a 1980s punk rocker from San Francisco named Klaus Flouride is just as important to me as the lyrics themselves.
Stories are more than just words.
On my leg
Supposed to beg
Daily to the filling station
Love and War
If you like this column, I hope you’ll check out some of our other endeavors at Love is a Fetish.
Semper Bellum: A podcast about war
Love Yourself!, a newsletter about self-care and being human
Check out more More Than Meets the AI here!