Otter's AGI test.
Apologies to Turing, Gödel and Minsky.
While am I mostly positive about AI and its impact on our little corner of the tech industry, I think we are a long way away from genuine General Artificial Intelligence. LLMs will not get us there. So it will need to be some other form of AI, which we probably don’t have yet. My bet is that we won’t really achieve it in my lifetime. It seems that a bunch of folks far smarter than me are thinking the same.
“The majority of respondents (76%) assert that “scaling up current AI approaches” to yield AGI is “unlikely” or “very unlikely” to succeed, suggesting doubts about whether current machine learning paradigms are sufficient for achieving general intelligence.”
See also Gary Marcus’ recent note. I’m not convinced we even really understand intelligence, consciousness and the mechanics of the brain well enough to know what to measure.
We probably aren’t such a long way away from the self interested benchmarks that the various manufacturers of Foundation AI propose as General Artificial intelligence, so I’m proposing my own alternative.
AI, build me a perfect fully automated German payroll.
As I usually do, here’s a tune or two.
One of my favourite German Bands.
As I’m just back from Vegas today we should have some Elvis.
Haha, I will do my best to make this the true definition of AGI.
Thank you Thomas Otter. I often find your take interesting and informative. I know tech enablement in people operations is here to stay and I am mostly okay with that. But I remain highly skeptical of specific applications for AI "agents". I am certain corporations are working overtime to figure out ways to replace humans. Yes, this has happened before with the Industrial Revolution. But this will be an entirely different order of magnitude. I worry about a world where there are not enough meaningful jobs. People need purpose not just pay. One thing for sure, Die Toten Hosen is fantastic! Thanks to you I found a new band!