BEN TSAI⋅BLOG

writing words on the internet since 00

AI: has the 🐎 left the barn?

ai has had its moments and then some this year and the hype continues to rise. everyday conversations with aunts and neighbors and shopkeepers inevitably include the topic. i've been doing some reading into the topic and (surprise!) i'm forming an opinion.

initially, my inquiry was anchored in my belief that humans are created by God and made in the image of God. so any definition of intelligence that infringes on this tenet is a no-go. so why did it seem like ai "experts" were talking about skynet becoming self-aware and obliterating humankind? (btw, i know how our story ends—it's not sentient ais).

turns out it's not all the experts and maybe those pushing for the fantasy narratives should be ignored and we should listen to people who are currently experiencing harms and exploitation at the hands of the rich. (I vehemently disagree with the part of lecun's analogy about God not existing, but i agree with the rest of his point:).

so coming back to chatgpt and the like—lots of folks are now using this technology. some are deeming it the next milestone in the software ui revolution. what i am concerned about is, the mental model that everyday users have is wildly divergent from how large language models (llms) work.

llms are not designed to tell us what is true. they don't lie or truth-tell or hallucinate. they output information that sounds good and coherent. relating to them as agents with reason is dangerous and harmful. couple that with the fact that these major LLMs players are not transparent about the data they are trained on, and that it is perhaps impossible to secure the data of these models, we're headed down an unnerving path.

the problem is very human. we want to believe. we crave relationships. (the bot love podcast series shows how deep and profound this can get). no doubt that many of us finding the tools useful. but what is happening when we engage with these bots seeking truth and answers and relationship? i feel like the rug could get pulled out from under us.

and yet all the big players (microsoft, google, openai) are heavily marketing these tools as agents you can have conversations with and find answers. a whole generation of companies are currently building on top of these platforms promising to help make business decisions, improve mental health, take users out of user research, and cure cancer.

i'm an even bigger ¯\_(ツ)_/¯ on ai now.


Recent posts