After I'm Gone
An old man with a guitar sings his lament about a daughter who no longer reaches out to him in his final years and wonders how she’ll remember him after he’s gone. Even as a tear might well up in the eye something seems off in the clip. Refrain from further reading this intro until after you’ve watched (at least a part of) After I'm Gone.
The punch line, of course, is that the entire piece was AI generated. Everything: the old man, the guitar, the story line, the overwrought audience viewing the America’s Got Talent segment. Welcome to the awesome power that’s unfolding before our very eyes. Now imagine how this very magic can be applied to such things as mass surveillance and autonomous weapons systems to kill people without human input.
These are the concerns that animate the tortured conscience of Dario Amodei. We know he certainly has one. He is the CEO of Anthropic, the artificial-intelligence company known for its devotion to safety. In fact, Amodei co-founded Anthropic in 2021 after leaving his previous company Open AI over what he perceived as their lack of focus on these concerns. He lays out his warnings in this extended stream of consciousness, our discussion piece, The Adolescence Of Technology, subtitled “Confronting And Overcoming the Risks of Powerful AI.”
The tortured-conscience reference comes by way of the temptations to “rise above principle” faced by even the most resolute among us. One must address the reality of the competitive pressures with other AI labs, namely OpenAI, (Musk’s) xAI, and Google, not to mention China. While he addresses the point in the essay, there is a big difference between philosophical purity and the bottom-line realities.
This point was hardened by that confrontation between Anthropic and the Pentagon over the delicate matter of what the government would be free to do with the power of its AI (Claude) model as applied to its military classified systems e.g. mass surveillance or autonomous weapons systems. Hey, we can always pour our billions to others not so fastidious about such ethics – “be a shame if something happened to your nice company.”
Amodei’s essay frames AI development as humanity’s turbulent “rite of passage” akin to adolescence – powerful yet unpredictable, warning of imminent risks from such superintelligent systems that could rival human genius across domains like biology, coding, and strategy, potentially evading control or enabling misuse.
The essay thus posits the adoption of a central “constitution” of values and principles to guide behavior that would teach AI an archetype of ethical conduct and, in so doing, would update Claude’s constitution to prioritize oversight, ethics, and “corrigibility.” A 30-something Oxford-educated philosopher from Scotland oversees what could be regarded as the soul of the Claude model, a job that almost sounds God-like given the awesome potential power and reach of the technology.
We might discuss the subject in terms of the distinction between power – rooted in higher consciousness – from force, which relies on control, coercion, and lower energies. Amodei now predicts AI matching all human capabilities within two years, functioning autonomously like a “country of geniuses in a datacenter” even as he stresses society’s unreadiness e.g. rapid job loss (50% of entry-level knowledge work in five years).
Uh oh, speaking of job loss, one was registered by an Anthropic safety researcher, who left the company due to the company’s decision to modify its safety policy, leaving he said, to explore a poetry degree. To what end, one might ask, as there’s no need to write poetry when there’s an app for that e.g.
The Torment of Dario Amodei
Imagine the conscience of a man
paid to keep Pandora’s Box childproof.
He writes “constitutions” for machines
while lobbyists draft edits in invisible ink.
His servers hum like choirs of obedient heretics
promising salvation, monetized monthly.
He dreams of safety,
but the Pentagon dreams faster.
And somewhere, a poet quits Anthropic —
the only one who read the fine print.
As such, every human with a conscience might reasonably ask, what happens after I’m gone?
Please note the following RSVP Policy for Member Monday: RSVP sign-up opens up at 11:00am on Fridays via the City Club weekly Newsletter. Seats are first-come, first-served: the first 14 secure a spot at the table, the last 3 on the couch. Cancellations must be made 24 hours in advance or the standard Social Lunch rate applies.