By Todd Sheets, Investment Banker, Economist, Author of 2008: What Really Happened
Doomsday Predictions
This past week, a 27-year-old researcher at Anthropic quit his
job in protest over fear that leading AI companies aren’t doing enough to
contain the risks of their rapidly emerging frontier models. After his departure, a senior executive at the
company seconded his concerns and added that he sees a 10% chance AI will wipe
out mankind.
Within a short time, the on-line post from the young
researcher went viral and sparked a fire in the political tinderbox that had
begun to form around AI. Anthropic’s
CEO, Dario Amodei followed up by recommending new global AI regulations to
curtail its development. Sam Altman, the
CEO overseeing Chat GPT publicly agreed with these recommendations.
Regulatory Capture
President Trump’s former AI Czar David Sacks has long
accused Mr. Amodei of promoting a narrative designed to help Anthropic achieve
what’s called Regulatory Capture, which occurs when large companies work with
government to limit competition. Indeed,
Anthropic and other frontier model developers would benefit substantially from
curtailing the growth of much cheaper open weight models that cost around 90%
less to operate and that are already
sufficient for a significant share of AI use cases.
The AI Genie and China
As Sacks and others have also pointed out, the AI genie is
out of the bottle and, all protestations to the contrary, there will be no
containing it. China (and other foreign
countries) will only use the idea of a global regulatory body to try and game
the system so that they might catch up and leap frog us. Given the economic and military stakes at
play this is an enormous risk, one that is quite likely substantially higher
than other doomsday scenarios.
President Trump, who shares this view, has boiled it all down
to: “Whoever wins AI wins”, and he’s right.
A Head Spinning Issue,
even for Elon
No less an authority than Elon Musk has been on all sides of
this issue. He co-founded Open AI as a
non-profit to help guard against potential risks and later sued its CEO (Sam
Altman) to prevent it from transitioning to a private company. Then, just this past July, Musk gave an
interview to The Economist where he said
AI is developing too quickly to be contained, in all likelihood it will be
enormously beneficial to mankind and, therefore, we might as well just sit back
and enjoy the ride.
But then this past week, Mr. Musk seemed to do another about
face when he backed Amodei and Altman’s calls to curtail AI development.
What should we do?
My thinking on AI has not changed since I first wrote about
it.
First of all, don’t panic.
Many of the doomsayers are the same people who, not that long ago, were
predicting that AI would soon create massive unemployment. So far, the evidence shows that just the
opposite is happening.
Further, statements from industry insiders that put a 10%
probability on total extinction are totally irresponsible. They have no idea how this will actually play
out, and their probability estimates only add a false sense of certainty to
their claims.
The fact is, the Genie is
out of the AI bottle and the potential economic and military benefits are so
enormous it will not be put back
in. Ceding leadership to China or any
other foreign country or actor would be a huge mistake, one that could
materially alter our future.
Many have talked about a regulatory body of experts to
oversee and control AI – but this is a false promise. The only people with the knowledge and
expertise needed to oversee this fast developing and cutting edge technology are
the leaders at the AI companies themselves.
In fact, no one has more to gain or lose, than these leaders. Whether they like it or not, it is their
responsibility to develop safe programs that benefit their clients (who are
ultimately paying for its development) and
to develop the systems that can protect their products and their clients
from hostile systems created by bad actors.
If they don’t think they are up to this job, they should
cede control of their companies to leaders who are.
AI companies may well need to pause development of certain
models as they take the time to create protections for their current and
prospective clients. But ceding this
authority to ill-advised political bodies would inevitably lead to either
uninformed regulations with dangerous and unintended consequences, regulatory
capture, or systems that are intentionally gamed by foreign actors seeking
their own benefit. Further, there is the
risk that too much government control will lead to uniform AI models trained
with biases that reflect the views of the regime in power, just as we saw
during Covid and as we see on a daily basis in authoritarian countries.
Modern history is filled with failed doomsday prophecies ranging
from nuclear annihilation, to warnings that an inevitable population bomb would
lead to mass starvations, to fears that Y2K would cause computer dysfunctions
that would cause to airplanes to fall from the sky, and that global warming
would doom mankind. Each of these
theories was supported by no shortage of highly educated “experts” and other
influential observers.
For nearly 6 million years now, mankind and our ancestors
have not only managed to survive, but to thrive. Today’s living standards are far superior to
what any prior generation has known and there is every reason to believe that
AI will help produce the next major leap forward in human flourishing.
Put simply: all of human history is on our side.
No comments:
Post a Comment