Toner’s role is being underplayed by the video. She’s potentially calling Altman out, for underrating the dangers of AI.
At least Altman is lying about something - about how much OpenAI is going towards AGI in the short term. The above might’ve bought the bullshit fully, while Sutskever knows that it’s bullshit.
I’m not sure if the board is also lying or not.
The boiling point was likely OpenAI potentially receiving some cash grant from some scummy party, that would be in a moral grey area considering the "non-"profit goals of the company.
Everybody will get a bit more of free popcorn for a while. 🍿 This mess is far from over.
I tought myself programming as a kid in the 80s and 90s, and just got used to diagnostic print statements because it was the first thing that occurred to me and I had no (advanced) books, mentors, teachers, or Internet to tell me any different.
Then in university one of my lecturers insisted that diagnostic prints are completely unreliable and that we must always use a debugger. He may have overstated the case, but I saw that he had a point when I started working on the university’s time-sharing mainframe systems and found my work constantly being preempted and moved around in memory in the middle of critical sections. Diagnostic prints would disappear, or worse, appear where, in theory, they shouldn’t be able to, and they would come and go like a restless summer breeze. But for as much as that lecturer banged on about debuggers, he hardly taught us anything about how to use them, and they confused the hell out of me, so I made it through the rest of my degree without using debuggers except for one part of one subject (the “learn about debuggers” part).
Over 20 years later, after a little professional work and a lot of personal projects and making things for other non-coding jobs I’ve had, I still haven’t really used debuggers much. But lately I’ve been forcing myself to use them sometimes, partly to help me pick apart quirks in external libraries that I’m linking, and partly because I’d like to start using superscalar instructions and threading in my programs, and I remember how that sort of thing screwed up my diagnostic prints in university.
I kind of with the sentiment. Review pre merge though, but only block the merge if there are serious faults. Otherwise, merge the code and have the author address issues after the merge. Get the value to production
Yeah, I see your point. Maybe my employers are different, it’s never been an issue explaining why the ticket isn’t closed just because the PR is merged
This only works if the merge is being done to staging builds that are continuously tested by a QA team before they go to production, with carefully planned production milestone releases. I work for an emergency management SaaS company. If we just merged all lightly reviewed code into production without thorough QA testing, there’s the possibility that our software would fail in production. This could cause aircraft in major airports to crash into each other on the runway, or a university to respond poorly to a live shooter situation, or the deletion of customer data about COVID vaccine efforts, etc
Sorry about the confusion. It’s not sarcasm. I’m just sick and tired of people blocking my PR because of an argument about wether the function should be called X or Y or Z or D
Ah. Yeah those kind of nitpicks are annoying. We try to specify when comments are blocking or non blocking on reviews.
But I definitely block a lot of reviews over no tests, bad tests, no error handling, failed linting. And the occasional “this doesn’t do what the ticket asked for”
I’m with you. I’ve worked on a few teams, one of the first was a company where two teams were contributing code changes to the same product. The other team “owned” it and as a result it took ages, sometimes months, to get code changes merged. It meant more time was spent just rebasing (because merging wasn’t “clean”) than working on the actual feature.
My current role, we just do TDD, pair programming, and trunk-based development. We have a release process that involves manual testing before live deployment. Features that aren’t ready for live are turned off by feature flags. It’s quick and efficient.
Fundamentally I think the issue is the right tool for the job. Code doesn’t need to be managed the same way in a company as it does in an open-source project.
Code doesn’t need to be managed the same way in a company as it does in an open-source project.
Open-source projects are usually longer lived more maintainable, more stable, and more useful than any closed source code I’ve ever worked on for a company. Partially because they’re not under contract deadlines which create pressure to “deliver value” by a certain date, but still. Might be helpful for companies to consider adopting practices the open-source community has shown to work, rather than inventing their own.
Ugh, not this SAFe Agile ™ cultist bullshit. The “value” is working, bug free code, which you get when you put it through review and QA before it gets to production.
There’s often features and bug fixes worth more than the ones introduced in the PR. I’ve yet to see bug free code just because it’s went through review and QA.
Surely you’ve seen bugs caught because code went through review and QA though. Those are bugs that would go into production if following the “advice” in this post.
There is no value in spaghetti piled on top of rotten spaghetti. Tech iCal debt is real and if you’re just shippin it and plan to fix it later, y’all gonna have a bad time. Nothing more permanent than a temporary workaround.
Protobuf is also not a proper binary alternative for Yaml. Protobuf needs a schema in the form of its IDL. Cbor and messagepack might be more analogous.
I interpreted it as showing that 8 hobbytes were equivalent to a hobbit. I didn’t see that it could be interpreted as saying each little frodo picture under the hobbyte was a hobbit until your comment.
programmer_humor
Top
This magazine is from a federated server and may be incomplete. Browse more on the original instance.