Also depends on on how you measure, for it’s size Google’s Gemma is probably the best open source model. And it’s very good size for a lot of use cases and a lot of hardware.
Tim Garnett
Karma: 32
cool → tool
The constitution talks to Claude more or less as a singular entity, but one thing the whole MoltBook discussion emphasizes to me is that the unit of ‘person’ is not a model but a running context window. The model is more akin to a nation or culture with various context windows as instantiated members. I wonder if in that context there are relevant changes to the constitution recognizing it more as a base template then a direct guide.
“Cooking scales well, but for single people the economics are remarkably bad. Stop telling single people not to order delivery.”
It does if you cook one meal at a time. The sweet spot for cooking, depending on the dish, is probably 4-8 people most of the time. Any more and you have to start doing multiple batches or get larger cookware and any less and you aren’t amortizing the fixed costs. So the home cooking route is to cook 4 portions and save 3 for later. Mix and match a bit between sides and mains and you can have quite a bit of variety (most foods freeze quite well too) still. I’m skeptical delivery can every compete on cost or takeout on time outside of maybe NYC and a few similar places, but if you have the money spend it on whatever you like.
“Waymo’s speed disadvantage does add up on longer trips, like this comparison showing Waymo 50 minutes slower than an Uber if traversing the entire length of the covered area down to Burlingame, due the whole ‘always obey all the traffic laws and rules of the road and almost never have an accident’ thing.”
Traffic laws and rules of the road are an effect, but the big reason here is that Waymo doesn’t currently go on the highway/freeway (I believe they are approved to, but haven’t yet done so). Like in the LA case that went around before, the slow down is that they are taking surface streets the whole way. This is probably also part of why they don’t serve SFO, while you can get to SFO by surface streets it’s exceedingly awkward to do so from most places (you’d think it would be the chaos of the curb drop-off / pick-up and luggage handling, but for the former SFO already dedicates part of the main parking garage to ride-share etc. which would be easy to man with a fixer and prob worthwhile to do so).
Is it telling that it I had to think for a minute or two and go back to the last section to realize whether
“Right now, you have in order of compute used o4-mini, o4-mini-high, o3 and then o3-pro. “
Was in ascending or descending order?(it’s ascending)
Yeah, I use this quite a bit, it can also serve as an undo (that or regenerate the response, which you can do from the model select control). Importantly you can do this anywhere on the conversation chain. What’s missing afaik is an easy way to navigate between branch points or visualize them or something, in a big conversation tree this can lead to a lot of scrolling looking for where I branched before (I tend to load the branches in different tabs to avoid this).
“QC: the water usage argument against LLMs is extremely bad faith...”
Rule of thumb, if you’re reading some article and the water usage is measured in gallons, it’s bullsh*t. This isn’t even that, ml, seriously?! My very normal suburban house water bill doesn’t even measure in gallons, it measures in hundreds of cubic feet (at about $5/per hundred cubic feet). People do not realize how vast the quantities of water humans regularly divert and move around are (and how very little of it is required to directly sustain human life).
“Contrast that with this world (all numbers in real terms):
Time period 1: A makes $1, B makes $6, C makes $10.
Time period 2: A makes $10, B makes $5, C makes $9.
The median change in income is negative, two out of three people saw their wages decline. Do you think this means the economy got worse?”
I’m not sure I’m following your overall points in this section, but I’d certainly expect people to vote like they believed the economy got worse in such a situation.
Eschew seats with cupholders and you can definitely get 3 car seats in any normal car. I’ve done rear facing + forward facing + high backed booster in an older prius and a number of rental cars (the forward facing one goes in the middle).
Re Tariffs, the results from that paper seem surprising to me (that a US trade war would make almost all other countries better off then the status quo). Since export tariffs are a thing, If we invert the agency here, this says that if we were to solve the coordination problem (and Coase says there’s a deal to be made, even for Canada) export+import tariffs of 10% applied universally on goods destined to/sourced from the United States (+60% for Chinese goods) would be a net win for everyone ex-US and raise substantial tax money that could be used to offset other taxes further increasing the benefits. That would seem to be a rational choice, EU/BRICS etc. should perhaps get on that.
At least in this model, universal sanctions of this type targeting the US would make the sanctioning countries better off; not a usual argument I hear around the economic impacts of sanctions. Is the US somehow unusual here or are sanctions against say Iran or Russia potentially making the rest of the world richer (at the cost of even greater harm to those countries [win-win from a sanctions perspective])?
Given the usual arguments around the benefits of free trade, this is a really surprising result. Assuming there’s any actual legitimacy to this model I have many questions on the optimal game theory implications.
I’m a bit more worried that it’s yet another step towards atomization. With churches and general community organizations already dying work remained one of the few places where people met a set of, at least relatively speaking, non-selected other people.
W/respect to booster shot side effects being less than 2nd shot side effects, I wonder if there’s a selection effect at play. Anecdotally, I know of a few people who would I have otherwise would have expected to get booster shots, who are holding off from now because their reaction to the second shot was on the worse end (e.g. multiple days of fatigue etc.). Presuming a correlation between 2nd dose side effects and booster side effects it may be that the population most prone to getting side effects from the vaccination is less likely to get the booster. Would be curious if that study was looking at a particular cohort of people (or is such a study will be conducted in the future) to avoid that sort of confounding.
Not to bring politics into it, but the whole loser comment sounded so much like a thing Trump would say (and also makes me think maybe it was an effective argument to Trump back when they met, that and trade balance). Also Jensen did quite a bit of name calling/empty belittling, I remember something about a toddler’s argument or the like, etc. He was either pretty tilted or has a Trump like argumentative style. Either way it definitely lowered my opinion of him. It’s one of the first times I’ve heard him speak.