phillypham

They removed the docs and support for it https://github.com/googleapis/python-genai/commit/af3b339a9d....

You can see the thoughts in AI Studio UI as per https://ai.google.dev/gemini-api/docs/thinking#debugging-and....

Depending how you count, parent comment is accurate. Hardware doesn't just appear. 4 years of planning and R&D for the first generation chip is probably right.

Could some disgruntled employee offer proof of seeding and collect a whistleblower reward?

I wonder if this also applies to people that live urban environments that don't drive and have to navigate walking or using the subway.

From my anecdotal evidence, it does seem that the average elderly person in NYC is way more active and social than an elderly person in the suburbs. But of course, it could be that people that live in cities self-select.

There is a lot of snark here. My non-cynical take is that they are aping the practices of top AI labs. OpenAI, Google DeepMind, Anthropic, and many other AI startup darlings have a culture of in-office work. People in non-hub offices are encouraged to travel a lot.

Yes, it will take a lot more than RTO to create an innovative culture like paying more for one, but one can reasonably hypothesize that working physically together is a necessary but not sufficient condition.

Yes, but the ability to identify prodigies and give them the resources to realize their full potential is part of an education system, too. I don't know the full details of the IMO participants, but the team is geographically diverse.

Of course, another aspect of the education system are the resources given to the average student, and I don't think there is much debate that the US could do better here.

At the very top, the US excels at math. We consistently place at the top or near the top in the IMO for instance (https://maa.org/news/usa-first-at-imo/). Yes the team is largely children of immigrants, but they are Americans, too.

Initial versions were one-click. The attack became more sophisticated and became zero-click.

See https://en.wikipedia.org/wiki/Pegasus_(spyware)#Development_... for timeline.

See https://en.wikipedia.org/wiki/Pegasus_(spyware)#Saudi_Arabia for the iMessage version.

It used to be possible to break into iPhones by sending just a text message without the target clicking on anything.

The only thing that kept this under control was there was an agreement to not target US-based numbers and the exploit was expensive.

Reference: The Battle for the World’s Most Powerful Cyberweapon https://www.nytimes.com/2022/01/28/magazine/nso-group-israel... and https://en.wikipedia.org/wiki/Pegasus_(spyware)

https://developers.googleblog.com/en/tensorflow-lite-is-now-... tries.

https://github.com/google/aqt is more explicit and preferable IMO.

Neither are as user-friendly as what Torchao has presented here.

As an American with Southeast Asian immigrant parents, they will live in a way that most Americans would find intolerable. Whole families in a 1 bedroom, very long commutes, taking buses, and living apart from their children (CPS, I know).

To be clear, I did not grow up like this, but I know many that did.

This is just (neurotypical?) human programming to be considerate of others. The driver feels this, too. I'd assume that the CEO of Uber Dara can manage his emotions and boundaries, well. But he mentions

Some experiences made him feel slighted, such as when riders discussed personal problems and company secrets on speakerphone, as if there was no one else present.

https://archive.is/2023.04.17-151927/https://www.wsj.com/amp... (WSJ article)

The blog post (https://ai.meta.com/blog/segment-anything-2/) mentions tracking as a use case. Similar objects is known to be challenging and they mention it in the Limitations section. In that video, I only used one frame, but in some other tests even when I prompted in several frames as recommended, it didn't really work, still.

Really cool. Doesn't really work for juggling unfortunately, https://sam2.metademolab.com/shared/fa993f12-b9ce-4f19-bb75-...

Awesome work environment for one person can be not ideal for another.

Pretty much all the top AI labs are both intensely competitive and collaborative. They consist of many former IMO and IOI medalists. They don't believe in remote work, either. Even if you work at Google DeepMind, you really need to be in London for this project.

You can break your sequence into two parts. One part goes through the encoder and the other goes through the decoder, so each token only goes through one transformer stack.

It's not too uncommon. I started off working with Angular and Java. But I studied math.

It depends on what type of role you want. If you'd be happy building the application layer and doing prompt engineering, just build applications that call LLM APIs.

If you want a research position at the top labs, the interviews really are actually passable by people without PhDs. They are really focused on having strong fundamentals. I've seen people make this leap but it can be years of preparation. Like actually reading textbooks, implementing low-level details like backprop, re-implementing papers, and doing non-trivial personal projects. Essentially, you're self-studying a Masters degree. Blog about it. Post about it here. I've found people to make this transition just generally love learning.

Mostly true in my network with the exception of people working on LLMs. They are motivated by nearly religious fervor that AGI is right around the corner. But I expect that to die down soon.

We support Google Pay and Apple Pay now, https://support.google.com/waymo/answer/14160746?hl=en.

Instagram shows posts from Threads in the feed. I accidentally click on these all the time.

I don't think money is entirely the reason that doctors don't serve underserved populations. Many states with underserved populations like Missouri already pay more due to supply shortages [1].

The reality is doctors have the ability to practice anywhere and tend to choose desirable places to live like NYC, LA, etc because they are humans with wants and needs, too. Anecdotally, living in NYC, many doctors and dentists tell me they could make more elsewhere, but they love living in the city.

[1] https://comphealth.com/resources/physician-salary-report-202...

In ML/AI space, at least some people thing analog computing is the next big thing, https://www.forbes.com/sites/karlfreund/2021/09/23/ibm-resea.... It hasn't taken off, yet, though.

Maybe it feels that way in the Bay area because the only companies are tech companies?

In NYC, it seems that tech is the only industry that allows remote work. Finance in particular is requiring employees to show up.

Aim for big tech companies. They regularly hire engineers that can pass coding challenges but otherwise have no experience or CS degree. I was hired at Google this way and many of my colleagues have similar backgrounds. It's a bit harder in this environment with hiring freezes, admittedly.

If you're more interested in startups that seem to require more practical experience, you'll probably need to do some personal projects or contribute to open source.

Unlocked link: https://www.nytimes.com/2022/10/09/us/california-high-speed-...

This is surprising to hear. I have a similar background, albeit my GPA was not quite as good and I terminated by mathematics education with a master's. I work with lots of PhDs that have transitioned from other STEM fields.

Are you good at math competitions? Competitive programming is basically the same thing. This is how I transitioned in the field. If you can do these, interviews will be a breeze.

I always felt Seattle got this reputation because we have so many transplants from California, particularly the Bay area and SoCal. Relative to there, it rains a lot.

Typically, these libraries are unusable for this purpose because you don't want to ship a Python interpreter with your video game.

I usually prefer to to rewrite my training step as a pure function, so the model weights are just inputs and the gradient updates are outputs.

You need to serialize your computation graph in some way, so it can be run in C++ or some other low-level language. TensorFlow is known for doing this well since it was original design goal of the project. Some of the other frameworks that originally targeted researchers make this harder. Most mature frameworks have some way of doing this now, though, and projects like https://onnx.ai/ may solve this in general.

It gets more complicated if your model has dynamic control flow, but you get the idea.

If you can make it into a FAANG/MAMAA/whatever flavor of the year acronym, you can make it into most tech companies (and still make a salary that's more than enough to live a dignified life and support yourself, and family if you have any).

Not actually true for early career SWEs, especially those that switched from a non-CS field. Because FAANG relies so much on coding puzzles, they are often the best option for someone that is self-taught without formal training. Since they often have a large recruiting pipeline, they are often the easiest places to get interviews without any connections, too.

I would also add that their ability to sponsor H1Bs is unmatched.

At Seattle and NYC farmers markets, I regularly pay $12 a dozen. The eggs are often sold out, so others are, too.