I had a crippling fear of flying and I still have anxiety about it. One way I got over it was by learning a lot more about aviation. Now I'm really not afraid of anything happening on a plane short of a. catastrophic failure like a wing snapping off or b. uncontrolled fire in the cockpit. Most other things are recoverable, and those problems are very unlikely.
The flaw that caused the fatal 737 MAX crashes was something that that better trained pilots would have been able to overcome. And this would likely be the case for most problems. The recent door plug blowout did not cause a crash, for example.
So I would not be afraid to fly on a 737 MAX from a major US airline over US territory. It wouldn't be my preference but I wouldn't be afraid to.
How is it opposed to OpenAI's goals to have a friendly company selling them chips instead of NVIDIA, which is, at-best, a neutral company?
Software is always more important than hardware. All the big players have access to NVIDIA chips today and yet only OpenAI has ChatGPT, proving the point.
OpenAI probably wishes someone would create competition to NVIDIA and this is Sam Altman trying to make that happen himself, since no one else seems to have been able to pull it off so far.
A conflict of interest would be OpenAI buying Altman's chips at inflated prices or something like that.
But if he makes a bunch of money selling OpenAI chips and OpenAI gets better/cheaper chips, that seems like pure win-win and totally free of ethical conflict.
A leader can't do anything on their own, they need people to lead. And those people deserve recognition and rewards. But in most cases there's no one more important than the leader. And thus, no one that deserves more credit than the leader.
I'm absolutely not comparing Sam Altman to any of these leaders, but just to illustrate how much vision and leadership does matter. Consider how stupid these statements sound:
"Jesus didn't build any churches, those were all built by brick layers and carpenters!"
"Pharaohs didn't build a single pyramid, those were all built by artists and workers!"
"Abraham Lincoln didn't free any slaves, he didn't break the chains of a single slave, that was all done by blacksmiths!"
"Martin Luther King Jr. didn't radically improve civil rights, he never passed a single law, that was all done by lawmakers!"
"Sam Altman didn't build ChatGPT, he didn't create a single ML pipeline, it was all done by engineers!"
It's a hard fact of life that some specific individuals play more important roles in successful projects than others.
If you want to get really basic: there's no OpenAI at all without Sam Altman, which means there's no ChatGPT either.
There are much larger armies of highly qualified scientists and engineers at Google, Microsoft, Facebook, and other companies and none of them created ChatGPT. They wrote papers and did experiments but created nothing even remotely as useful.
And they still haven't been able to even fully clone it with years of effort, unlimited budgets, and the advantage of knowing exactly what they're trying to build. It should really give you pause to consider why it happened at OpenAI and not elsewhere. Your understanding of the dynamics of organizations may need a major rethink.
The answer is that the CEO of OpenAI created the incentives, hiring, funding, vision, support, and direction that made ChatGPT happen. Because leadership makes all the difference in the world.
It's completely ignorant to discount all organizational leaders based on your extremely limited personal experience. Thousands of years of history proves the difference between successful leaders and unsuccessful leaders.
Sam Altman has been an objectively successful leader of OpenAI.
Everyone has their flaws, and I'm more of a Sam Altman hater than a fan, but even I have to admit he led OpenAI to great success. He didn't do most of the actual work but he did create the company and he did lead it to where it is today.
Personally, If I had stock in OpenAI I'd be selling it right now. The odds of someone else doing as good a job is low. And the odds of him out-competing OpenAI is high.
Wrapping every third-party library in defensive code doesn't make a lot of sense since this isn't the most dangerous kind of bug. To make a one example of a million: a third-party database library that corrupts your database is infinitely more dangerous and that there's no form of wrapping that will save you.
Bad code is bad code. Bugs are bugs. This issue is not a major source of problems in Go code.
Sounds like you may be confused thinking that a panic in Go is like throw in other languages? Third-party libraries should virtually never panic. Go has panic, and it's useful for a few things, but does not use them as exceptions are used in other languages.
This is a bit like worrying that your programming language won't prevent a third-party library from sending a SIGKILL to your program or that it doesn't prevent third-party libraries from delete your entire file system.
No general purpose language protects against malicious or stupid code.
My sense is that you probably could help them get their technical act together, if the team was otherwise in good shape, but that you're probably not the right person to provide the kind of comprehensive leadership they need.
What's required to fix this situation is for someone to manage up and down the hierarchy, and it doesn't seem like you're going to be able to pull that off in this organization. Maybe someone with more experience could make it happen but it's probably just a shitty company and it's not worth the effort.
The fact that the CTO and directors let this situation fester for years is a sign that they're clueless and/or disinterested.
Best alternative would probably to create a new team to replace the previous team's services one-by-one with well-operated versions. And then fire the previous team entirely. Probably not a very fun job.
I wouldn't expect them to be lying but there to be some data hiding in the "average" part of the ARPU. We often want the median or some other percentile, not the mean.
"We define ARPU as our total revenue in a given geography during a given quarter, divided by the average of the number of MAUs in the geography at the beginning and end of the quarter."
So that is about what I expected. I'm not going to dig into this but it could be very meaningful or extremely misleading. Interesting though.
I've used Digital Ocean (and many other hosting providers) for as long as most of them have existed. Most of my servers have been running nearly uninterrupted for many years. Yes, there will be a reboot or move every so often but the uptime is incredibly high.
The idea that single server is capable beat the reliability of a massively distributed system is counter-intuitive and yet usually it's the case.
The average distributed system is a house of cards that can come tumbling down if any one of a number of pieces fails. The average static server is a rock of stability, with very few failure modes.
That seems plausible but can you cite where these figures are from? Is this just a rough estimate from total numbers for the entire company of $REVENUE / $MAUS?
Most customers care about having the best of the available options. Rarely would any company deliberately choose to be behind where their competitors can be.
1. The way to run into undiscovered issues is to choose a completely custom firmware/hardware/software stack that almost no one else in the world is running.
2. Not sure where you're getting this from. There is almost always a price:performance calculation that results in current generation smashing the previous generation with server and switch hardware. Often this means not buying the flagship chips but still the current generation.
And a major reason to get off old generations of hardware is that they become unavailable relatively quickly. It's always easier to buy current generation hardware than previous generation hardware, especially a couple years into the current generation. This has nothing to do with chasing the latest hardware.
If you're not rapidly scaling it probably doesn't matter. But if you're still buying (and maybe even using) Haswell CPUs in 2023, you may be missing out in a big way.
A moderately large Haswell cluster is equivalent in power to a moderately powerful modern server.
Coupling requires more integration work, including writing and testing custom firmware. Oxide will be a tiny market player for a long time, even if things go very well. Are AMD and Broadcom really going to spend as much time helping Oxide as they do helping Dell? Of course not, Oxide's order volume will be a rounding error.
I'm sure they'll improve their processes over time but the lag will probably always be a non-zero value. Hopefully they'll be able to keep it low enough that it's not an important factor but as a customer it's certainly something one should consider.
It would be surprising if they don't run into some nasty issue that leaves their customers 6+ months behind on servers or switches at some point.
Coupling vs Decoupling is not some one-sided thing. It's a major trade-off.
One of the most obvious examples of the problem with this approach is that they're shipping previous generation servers on Day 1. One can easily buy current generation AMD servers from a number of vendors.
They will also likely charge a significant premium over decoupled vendors that are forced to compete head-to-head for a specific role (server vendor, switch vendor, etc).
Their coupling approach will most likely leave them perpetually behind and more expensive.
But there are advantages too. Their stuff should be simpler to use and require less in-house expertise to operate well.
This is probably a reasonable trade-off for government agencies and the like, but will probably never be ideal for more savvy customers.
And I don't know how truly open source their work is but if it's truly open source, they'll most likely find themselves turned into a software company with an open core model. Other vendors that are already at scale can almost certainly assemble hardware better than they can.
One can easily imagine the privileged son of a medieval samurai attempting to purchase a lemon-lime Shasta from a vending machine outside his castle, only to have his coin repeatedly drop into the return slot.
You're doing it wrong if you think programming for work is about the code being a means of expressing your inner soul. The most successful professional programmers derive their enjoyment from achieving business objectives, like making products that users love and improve the world.
If you want to express your soul in your code, do that on the weekend using whatever language you want. Trying to mix these motivations is a recipe for disaster.
Go is a complete and simple language, and this is a fairly objective claim.
Unless by "incomplete" you mean it could have more features. But C++ and Rust (and Java) have been piling on features for years, with no signs that they will ever slow down or stop, proving that they're also "incomplete" by this definition. IMO this is really just a result of there being too many cooks in the kitchen, and a lack of leadership committed to saying "no" to feature requests.
This is also why using Rust or C++ requires every organization to agree on a subset of the language they will utilize. Rust isn't as far along this path, but it's heading in the same direction as C++.
But with Go, every organization can use the complete language. That's how simple it is.
Go has been used by many organizations to build stable, large, and scalable systems that have been operating successfully in production for many years now. That's how complete it is.
The flaw that caused the fatal 737 MAX crashes was something that that better trained pilots would have been able to overcome. And this would likely be the case for most problems. The recent door plug blowout did not cause a crash, for example.
So I would not be afraid to fly on a 737 MAX from a major US airline over US territory. It wouldn't be my preference but I wouldn't be afraid to.