synchronousq

And what value has the current state of AI added to society in any meaningful way? Truly? Even as someone who understands the space very strongly, and has published several top-tier ML papers, I cannot help but to conclude the primary destination of the current tech sector and Silicon Valley is the ruthless exploitation of the rest of society.

Taking a sober look at the state of software, we observe a few things.

The services offered by modern software to users, as a whole, have remained largely the same over the past ~5 years. The state of software quality is in rapid decline, with enshittification and rent-seeking running extraordinarily rampant. Software security has been in the same disaster-state it has been for the past 20 years, where software resilience is in stagnation, governments and private institutions stockpile vulnerabilities, and security researchers and auditors can consistently find new vulnerabilities. The rest of American society outside of the tech sector is currently facing a standards of living nosedive, and clearly they have not benefited from the tech sector's financial proliferation in the AI space.

Realistically, I cannot help the feeling that we're headed towards a reality where the 4th amendment is dead, and machine learning models process everything about you to ultimately extract more from you. No privacy for you! No agency for you! Only indentured servitude, and constant fear.

I fully recognize my take is ahead of its time, but I concur that the systems-oriented point of view is our way out of this hell. Specifically, software should be conceived under the following ideals: (1) software should be as simple as possible, and provide its intended services with as little bloat as possible; (2) specifications of software should be as concise and simple as possible; (3) specifications should be should be expressive enough to capture security-relevant guarantees, e.g. cryptographic security properties; (4) proofs verifying that software satisfies its specifications should live intrinsically to the implementation, and should be as simple as possible; (5) proof-checkers should be verified. I feel the academic Formal Methods, Programming Languages, Systems, Security, and Cryptography communities, as well as the internet standardization community, are slowly converging to this ideology consensus, but I also think in other ways we are farther off than ever. With respect to these ideals, the "building" mindset that twitter has adopted is deeply toxic. And obviously Silicon Valley has their heads in the sand when it comes to this.

I do have faith the state of software (and society) will improve, but whether that future is compatible with the rent-seeking hyper-capitalist reality Silicon Valley and Wall Street have synthesized is yet to be seen.