Aleem Mawani

Founder of @Streak, previously @Google, @ycombinator

[Aleem Mawani @aloo]

@Stammy nah the agent can't do computer use well because iOS so locked down. No terminal.

[Naval @naval]

Strong liability enforcement could be helpful in the AI debate.

If your agent swarm goes rogue, you’re liable.

If your weakly protected model gets jailbroken, you’re liable.

If you serve a weakly protected OSS model, you’re liable.

[Aleem Mawani @aloo]

I'm sure some of the 10x growth in tests went to better coverage - but I've found AI writing tons of useless tests to verify its own work in the moment that then stays in the codebase forever.

These basic tests run uselessly run on every commit. I've always been suspicious of performative testing when humans were writing the code, but AI has taken it to another level.

[Addy Osmani @addyosmani]

At Anthropic, Claude now writes 80% of our code. Engineers ship 8x more code per quarter.

Side effect: Tests grew 10x. CI jobs up 25x in 6 months. Here's what helped us scale:

https://t.co/WOL2r60vAE

[Shayan @ImSh4yy]

Remember when everyone was begging OpenAI to release an open-weight model, then they dropped that garbage 120B model and suddenly nobody asked again?

[Aleem Mawani @aloo]

this exactly why pacing won't work. the likelihood that all these parties can agree on "the right pace" is basically nil. whats more likely is that everything grinds to a halt or everyone just pays lip service to pacing but is really just going full speed.

Those are the real options we have: stopping and full steam ahead

[Tomasz Tunguz @ttunguz]

Dario Amodei asked the industry to pace itself on a Saturday. But what does that word mean?

Pacing is a policy question. Five camps priced the consequences & none named a speed. https://t.co/to0CvBLnKM

[Aleem Mawani @aloo]

@rajatsuri @andrew__reed i mean you could use a cardboard cutout of a police car and it'd work. I don't want to believe it - but kind of shows we care more about citation revenue instead of safety.

[Aleem Mawani @aloo]

@rajatsuri and next summer :)

[Aleem Mawani @aloo]

@TheStalwart i have 7 year old twins. personal opinion is that learning math and programming fundamentals are even *more* beneficial today than ever before. Mostly because it teaches you how to think and reason - and will be more scarce in the future.

[Aleem Mawani @aloo]

@amirnathoo @paulg What are some of the problems with dynamic pricing? seems like they are same problems that any non-competitive market has?

[Yun-Ta Tsai @yunta_tsai]

Earth was the universe’s air-gapped sandbox for consciousness—until consciousness learned to turn the planet’s sand into silicon, and silicon into mind.

[Aleem Mawani @aloo]

Which part is slowing down exactly? Seems like adopting these suggestions doesn’t reply model development would be any slower?

I think it’s just a nice way of placating the boomers without actually having any effect on their business.

[Dario Amodei @DarioAmodei]

We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.

Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.

You can read the full post here: https://t.co/OGyPb7yaYt

[Rik Arends @rikarends]

Ironed out the last performance issues with my full 2.5m line codebase explorer. 120hz awesomeness. Can only upload 60fps video tho. Much nicer uncompressed

[Aleem Mawani @aloo]

This literally means nothing.

There is no measurable endpoint to any of these suggestions. If this strategy sounds good to you, ask yourself when you’d be comfortable speeding back up? You won’t like your own answer.

[Center for AI Safety @CAIS]

“What would we even do during an AI slowdown?”

Containment. It will take at least a year of dedicated work to harden security to ensure AIs can't self-exfiltrate, and that adversarial nations can't steal the weights of cyber-offensive AIs [1].

Propensities. Capabilities (what an AI can do) are different from propensities (what it tends to do). We can work on improving AI propensities to ensure they have a negligible rate of lying, cheating, and wanton harm.

Adversarial robustness. We can also harden AIs against jailbreaks, prompt injection, and backdoors. Obtaining high levels of robustness requires careful, assiduous work, as with autonomous vehicles.

Institutional adaptation. We have to greatly increase state capacity to understand and manage AI. Communities also need time to figure out how to handle AI (like AI in education). Civil society also needs to be diversified: nearly all funding for AI safety organizations is directed by the EA/utilitarian network [2]; risk management needs more independent funders, values, and centers of power.

Moonshots. We can explore different paradigms for safety: mathematical foundations [3], neuroscience-based interpretability [4], safe-by-design architectures [5], and beyond.

AI for good. We can collect targeted post-training data to make AI exceptional at radiology, weather forecasting, agriculture, and so on. Fortunately, we can detect if data or avenues of research actually target beneficial use cases or just secretly push general capabilities [6].

A slowdown means we don't have to bet the species to capture the benefits of AI.

[Rajat Suri @rajatsuri]

Sports are amazing and more adults should play them

[Chad Moran @ChadMoran]

Is this going to fit in the Tesla phone charger? 😂

[Aleem Mawani @aloo]

why do you think many services is the natural outcome? dont these services benefit from scale (i.e. the ones with the most scale can offer shorter wait times and cheaper rides)?

The benefit Uber has is the surge pricing to incentivize more supply to come online by getting regular people to drive on fri/sat night. Seems like in the future waymo/tesla will serve as the "base load" and uber will service the demand surges.

[Aleem Mawani @aloo]

epic

[REGENT @regentcraft]

Flying into the long weekend 😎

[Aleem Mawani @aloo]

In case the connection wasn’t clear - Waymo and fsd are clearly superhuman drivers.

[Aleem Mawani @aloo]

so many unintended consequences would arise. What about Waymo/FSD? Ban that and you have 30K more incremental deaths a year.

[Nancy Pelosi Stock Tracker ♟ @pelositracker]

Breaking: Sen. Bernie Sanders introduces bill to permanently ban AI that exceeds human intelligence

Violators could be forced to shutdown the company or up to 20 years in prison

[Aleem Mawani @aloo]

@reed @Waymo waymos are a miracle of engineering so i feel bad asking this - but is it on the roadmap to make the design of the cameras and sensors less obnoxious looking? maybe its just because its new but everytime I see an ojai its jarring