This hot and messy blog post was written by a moderately cited worry wart of a machine learning researcher working for Google, who has written/co-authored about 130 papers on machine learning! He is also one of the researchers I am following in my studies of AI/machine learning.
How does it feel to be turned into a "paper clip" or what is a "toilet tank mechanism" you have to "jiggle just right"? The Walt Disney fantasy of an AI researchers!
See also my very recent post touching on similar issues regarding AI.
It is not up to Jascha or his compatriots "to get to wonderful outcomes" whatever this might be!
Elitist PhD experts like Jascha will also not get to define what is misaligned with human values!
As far as his exaggerated fear of "a subtle misalignment might pose an existential risk to humanity" is concerned, he might be too young to realize humanity managed so far very well to live with ABC weapons! The suspected lab leak of SARS-Cov-2/Covid-19 also did not kill all humans! However, still existing totalitarian dictatorships like in Russia, China, or Iran pose a serious risk to the potential of AI being used for malicious purposes.
The young man is also afraid of superintelligent machines! Scary or is it more like alarmism and hysteria? Is it not more likely that superintelligence will benefit humans too by becoming themselves superintelligent like in symbiosis?
"Many machine learning researchers worry about risks from building artificial intelligence (AI). This includes me ...and we will need to work hard to get to the wonderful outcomes ...
One popular AI risk centers on AGI misalignment. It posits that we will build a superintelligent, super-capable, AI, but that the AI's objectives will be misspecified and misaligned with human values. ... a subtle misalignment might pose an existential risk to humanity ...
I am extremely glad people are worrying about and trying to prevent negative consequences from AI. ...
I am extremely glad people are worrying about and trying to prevent negative consequences from AI. ...
An accidentally misaligned superintelligence which poses an existential risk to humanity seems about as likely as any other specific hypothesis for the future which relies on a dependency chain of untested assumptions. ..."
One popular AI risk centers on AGI misalignment. It posits that we will build a superintelligent, super-capable, AI, but that the AI's objectives will be misspecified and misaligned with human values. ... a subtle misalignment might pose an existential risk to humanity ...
I am extremely glad people are worrying about and trying to prevent negative consequences from AI. ...
I am extremely glad people are worrying about and trying to prevent negative consequences from AI. ...
An accidentally misaligned superintelligence which poses an existential risk to humanity seems about as likely as any other specific hypothesis for the future which relies on a dependency chain of untested assumptions. ..."
Credits: Synced Global AI Weekly
Following is a very nice cartoon! 😊 Imagined corner cases, oh well! Nice exercise!
Figure 1: The space of intelligence and coherence. Each corner corner represents an extreme of intelligence and coherence, and is labeled with an example of a machine demonstrating those attributes.
No comments:
Post a Comment