conjecTech

Hawaii is the only state that produces a significant fraction of its electricity from oil. Something like 40% of continental US electricity is already low-carbon(nuclear, wind, solar) and the bulk of the rest is natural gas, which produces significantly less emissions per unit of energy than diesel. From a CO2 perspective, it is almost certainly better to drive a full electric vehicle, and that will only become more true as we continue to transition production.

The table they pull that data from is annual and only has info through 2022[1]. It's not useful for saying anything about the year-over-year monthly number being discussed. The chart showing monthly sales of road petroleum show a consistent 5% annual drawdown, which is proportional to how quickly the fleet is transitioning[2].

[1] https://www.ssb.no/en/statbank/table/11561 [2] https://robbieandrew.github.io/EV/img/fuelsales.png

None of those factors are unique to this business. Somehow adults at thousands of other companies are able to handle this situation reasonably despite also having to pay rent and taxes.

I'm sure very few of the people whose continued work they were depending on for a sale would have stayed this long if they realized they were going to be treated like this.

Actions and comments like this are how you kill an ecosystem.

This is the worst take I have ever seen. This company just raised $250 million. That is more money than 99% of companies will ever have in the bank. Being a "startup" does not given you license to wantonly play with your employees' lives and wellbeing. As the quote from Dalton above shows, there are norms around these things which companies 1/100 this size regularly abide by.

This is not normal behavior for any company, and you should run away from anyone trying to convince you otherwise. The people responsible should be treated like pariahs.

I mean, there is the rather extensive "Sexual harassment investigation" section of his wikipedia page where he paid $1.4 million as part of a jury verdict in a case of sexual harassment.

https://en.wikipedia.org/wiki/Bobby_Kotick#Sexual_harassment...

Can anyone explain this from an RF/networking perspective? What frequencies does this operate at? How well matched are the antennas in most phones? How much more power are you using to be able to do this? What kind of modulation are you using for TX/RX?

Sounds like you live in Long Island. You seem to admit that if not for the local government putting restrictions on how people use their land, individuals would choose to build multi-family structures. It is a compelled choice, even if only for a minority of the population.

I'm referring to things like single-family zoning, minimum parking requirements, and the subsidization & forced construction of highways through city centers. The situation we have today isn't some spontaneous expression of consumer preference. Federal, state and local governments have made building walkable alternatives nearly impossible and prohibitively expensive.

That is not the alternative being discussed here. 80% of the US population is urban. We could not undo that without a huge degradation is quality of life. Cities are incredible engines for wealth creation. The alternative more or less imposed by law is suburbs that don't have access to communal space or nature.

Something like 80% of Portugal's solar has been installed since 2018[1], when solar was on-par with if not cheaper than any other form of electricity generation. And that's from aggregate perspective, the economics are definitely better for Portugal. There were PPA being signed in similar geographies for 3c/kwh that year. I wouldn't be surprised if the defrayed cost from lower fossil fuel usage just since the start of war in Ukraine has fully offset the cost of the installation. I'm not sure why you think this has been so costly for average citizens.

[1] https://www.pv-tech.org/portugal-installs-more-solar-pv-in-2...

There will definitely be induced demand. Most people don't just choose travel mode based on cost though. Travel time/convenience is the biggest factor, and AVs will probably not improve those significantly. The reality is that walking/biking/transit aren't competitive in terms of convenience in most US cities right now. If taking the train is faster than driving, people will do it. That is the fundamental issue that needs to be fixed to allow other options to be successful, and I view killing people's obsession with parking as the largest step in that direction.

I'm a big urbanist (that also worked on AVs for a couple years), haven't owned a car for almost all of my adult life, and view it as a step in the right direction for a few reasons.

1 - I think the impact of parking is actually more insidious and detrimental than roads by themselves. AVs will have high utilization of a much smaller pool of vehicles and remove the need for the huge parking lots we currently have. The land-use issues from parking are perhaps the largest impediment (together with the requirements/zoning) to the density which would allow transit to outcompete private cars.

2 - The centralization of wealth is one thing, but the real economies of fewer, newer vehicles will be a boon for almost everyone. The cost of the automobile operation is about 1/4 of the average uber fare. That is a lot of room for price compression. Perhaps more will be going to a large corporation, but there will be a huge consumer surplus as well.

3 - The reduction in the size of the car fleet will likely reduce the lobbying power of the automobile industry, which may in turn lead to changes to the urban environment(bike lanes, transit expansions) which will likely shift mode share away from cars.

If you do the math on that cycle count, you'll see its several times the warrantied amount. 1500 charges at 80% for a standard range M3 is over 300k miles. I think they probably picked that value to align with norms of other warranties rather than reflect any kind of real risk. They released some charts as part of their impact report showing 12% degradation at 200k miles, which lines up pretty well w/ the 1500 charges. [1] Most people drive about 10k miles/year, which would maybe be 50 cycles/year, and obviously the batter is unlikely to last 30 years.

[1] https://electrek.co/2023/04/25/tesla-update-battery-degradat...

The last time I looked into this the primary source of capacity degradation in EVs is from aging, not cycling. Most users could probably cycle their battery 50% more for a minuscule loss of overall lifetime/utility.

I wish _all_ students with that profile were automatically accepted. I went to GT and while I guess it's nice in some ways that its now considered an elite university, I much preferred the persona of admitting everyone and weeding out people out so that they could give kids without stellar resumes a chance.

Inflation isn't a scalar value, and all of this complexity is just trying to account for that. This thread is about whether an increase in effective income from prices going down is the same as your notional income going up while prices stay the same. My point is they're not, but feel very similar in most cases. I used an example with inflation rather than deflation since people understand that more intuitively.

Size is denominated in terms of some measure. You don’t eat dollars, and you can’t build a house out of one. GDP has doubled, but so have US prices. Does that mean the economy is actually stronger? Using PPP simply chooses to denominate GDP in terms of things other that fiat currency.

I've worked on plenty of hastily built houses from the 70s that are still standing half a century later and in relatively good condition. They may run up slightly higher maintenance bills, but I'd be surprised if they don't last to see 100 years. Houses today are undoubtedly better constructed. The quality of materials in the old days compensated well for the stress from poor environmental and thermal design, but you can probably get the same longevity by designing the house to not have that stress in the first place.

Congratulations! How do you plan to make money?

That's part of my point. There are 100s of organizations using it at scale, but it only needed to be trained once.

The more general the model, the longer the lifetime. And the most impactful models today are incredibly general. For things like Whisper, I wouldn't be surprised if we're already at 100:1 ratio for compute spent on inference vs training. BERT and related models are probably an order of magnitude or two above that. Training infra may be a bottleneck now, but it's unclear how long it will be until improvements slow and inference becomes even more dominant.

Capital outlays are tied to the derivative of compute capacity, so even if training just flatlines, hardware spend will drop significantly.

No, there isn't much nuance here. Accounting has notions of fixed costs and marginal costs. The startups you are referring almost all lose money on fixed costs, but sell things at per-unit economics which make sense at scale because they are below marginal cost/COGS.

Things like Uber are a typical dumping cases. For years, they charged below their COGS and eventual profitability depended on driving out competition and raising prices.

Sounds great to me. Especially considering you aren't paying any taxes on your returns when they come in the form of deferred costs. That's equivalent to a pre-tax return in the mid-teens. You won't find that in most asset classes these days, and definitely not with the risk profile of solar.

I assume you care about 2D vs 3D for different variables, no? I agree 3D, at least while H<<W, doesn't matter for clock speeds, but wouldn't it still play a significant role in the power characteristics, particularly things like gate capacitance? Genuinely curious. Going off of my memory of my Physics of CompE class ~10 years ago at this point.

Transistors at that scale are 3D rather than 2D, so the expected scaling ratios no longer apply. If you wanted double the density of a 2D transistor with 5nm features, you'd need to scale down to 3.5nm, but if you wanted to double the volumetric density in 3D, you'd only need to go down to 4nm.

How do you make meaningful compare to past generations with that being the case? You could then just describe it in 2D terms, and refer to your 4nm process as 3nm. Then all of the same rules of thumb still hold.

I assume this is responsible for at least some portion of the deviation and confusion, though maybe not all of it. Happy to be corrected if I'm wrong.

5 years is probably a fair current estimate, but that also doesn't come for free. The fans in DC servers move a ton of air, which itself further increases power use. I was thinking more on the order of 10-20 years.

Yeah, that video is fantastic. I wonder if the higher costs will place more of an emphasis on running chips at power levels that will gives better longevity. We can effectively halve operating cost and double lifetime of chips by reducing their power levels 20%. Seems plausible if we don't have technological obsolescence to force depreciation.

You get most of the economic value from the first few divisions. Going from 80% yield to 90% yield shaves 14% off your cost/unit. Going from 95% to 99% only saves you 4%.

I'm not an expert at how they combine chips. Like I said for AMD, they also wanted their units to be composable w/ small number of chips, so they basically have a die w/ a few cores and different die w/ memory, and I believe they have a proprietary communication mesh for connecting them. I think there is some considerable signal/energy overhead to communicating between chips. The cost of masks and interconnects is probably high enough to make a high number of diverse chiplets inviable, but I wouldn't say it's impossible in the future.

Chiplets aren't so much driven by a need for heterogenous processes as much as minimizing the yield impact of defects. Ian Cutress has a good portion covering it in this recent video: https://www.youtube.com/watch?v=oMcsW-myRCU.

In short, if your process produces 20 defects per wafer, and you can fit 100 chip on a wafer, you're going to end up with ~84% yield(ie, slightly less than 20% loss). If you are able to split that same chip into 2 equal-sized pieces and make twice as many, your yield is now above 90%.

AMD has also made them composable in such a way that they have to produce and stock far fewer ICs to fulfill all of their SKUs, which is also another fantastic benefit.

An EV is a huge battery on wheels. And typically it's far cheaper per kwh than dedicated solutions. Combining the two is a fantastic source of economies and convenience. I'm sure a lot of people in California would have appreciated the optionality of V2G this summer when there was a threat of blackout and PGE was paying $2/kwh for people to supply the grid from their local storage.