@seanluomdphd@AlexGodofsky So roughly 3 GiB of data, and the neural instruction set is comprised of roughly 40-80MiB.... so roughly 10-50 orders of magnitude more efficient than a frontier LLM. Right... It actually makes it far worse in terms of efficiency.
@Justagirldog@ryangrim Because Democrats finally want to win and it's better to have someone to the right of your Candidate win who is closer to you ideologically than the Republican Candidate? Isn't that obvious?
@synthshareai What makes 0365 all that much better than Workspace for like 99% of use cases? In particular on collaboration WS is significantly better IMO. ChromeOS is just a completely different market, not a direct competitor to macOS. Gemini is certainly valid though.
@_simonsmith Didn't happen with Cloud Compute, not sure why it'd happen here. Unless they have a reseller agreement it would likely be against the TOS. OpenAI may eventually have channel sales but if it's the same as cloud, you don't buy capacity for your end customer, you get a spread.
@MSMInsider@gbrew24 You mean like... A joint plan with Iran and other coubtries? With comprehensive controls? That has some actions performed by Iran and the other signatories to ensure compliance? We could call it... The Joint Comprehensive Plan of Action.
@staysaasy Not an engineer, I needed a tool for pipeline management for a whole team of people. Had a bunch of requirements around RBAC, data visibility, process flows etc.. Getting eng to build in our CRM? Months of time, low priority, high friction. Vibe coded myself in a few weeks.
@staysaasy No one is replacing enterprise grade sw with vibe coded tools unless they didn't need like 90% of the functionality. But there is probably a lot of low-hanging fruit out there that sits somewhere in between a spreadsheet and enterprise grade app.
@Chris_Worsey@the_P_God@karpathy You cannot backtest agents with any data prior to their knowledge cutoff. Agents would be contaminated with financial data and news of stocks in the backtest rendering it useless because you wouldn't have out of sample data by definition...
@JLima29426@bydylanlamb@TheZvi No you illiterate chud he's saying that Elon will gladly go along with whatever heinous shit he is asked to do, while justifying it as being "part of the system" so long as it protects his ability to achieve his own personal goals or interests.
@garrytan Why wouldn't all these bots be powered by gpt-10 or Claude like they are today? Who is going to be training their own models if you have God level intelligence available? Have you thought this through at all?
@FlexFutures00@paxtrader777@GovTimWalz "As a black man". Lol there's been illegal immigration since forever. No one is arguing against deportation of undocumented immigrants. It's the way in which it is being carried out that is the problem. Prior presidents deported more than Trump without sending an ICE army.
@paxtrader777@GovTimWalz Ah yeah... It's not the people who sent in an armed force to ask brown people for their "Papers" in the most aggressive manner possible that are responsible... It's the people saying it isn't right. I like your trading advice but this ain't it man.
@benitoz You're missing the point. Google proved you can design your own chip without NVIDIA and train a SOTA model with it. Others will too. NVIDIAs software licensing costs are prohibitive.
@levelsio@scotthill222 Much of the poor that get free money can't vote anyway - Children. But their interests are expressed via their mothers/father's/family's votes. Then there are the indigent - those who cannot work due to disability. Why should their interests not be taken into account?
@levelsio@scotthill222 Lots of people get "free money" through all sorts of distortions in the tax code and everyone votes in their own interest. Why are the interests of the poor less valid? Disenfranchising people because you disagree with their interests is wrong.