Seconds count for 911 calls, but really your odds are already bad if calling about...a heart attack. There's one study about non-runners having heart attacks during marathons due to road closures [0]. If they had a heart attack that day, they were 15% more likely to die within a month. Not good, but it's not that bad.
Going full SV utilitarian, I'm curious what's the net change in accidents between
(1) texting
(2) no texting?
I've read that texting is the equivalent of having 2 beers. Even "hands free" is distracting. I continue to see people sucked into their phones and oblivious that they're operating a 4,000+ pound machine.
>Well, you're picking extremes when AFAIK, it'll put the average person at the legal limit.
>One beer will start to impair you.
Thank you for illustrating exactly the problem. Impairment is a binary in colloquial usage. Statistically no average-median-ish person has ever been impaired in the colonial sense by one average beer. Any everyone knows this. Two average beers applied to an average person won't get you to the legal limit without aggravating circumstances (i.e. zero time to metabolize + empty stomach, or perhaps conflicting medication).
I will be the first to admit you can give a bunch of people one beer and detect statistically significant difference vs a control group or you can give one person one beer many times and evaluate against a baseline and detect a statistically significant difference. But statistical difference does not "impairment" in the colloquial sense make. And everyone knows this based on their own observed life experience, even people without experience should be able to deduce this by observing how the world behaves for if what you say were true, the way things work would be very different.
And by using the term impairment to describe/quantify the impact of one beer and then re-using that term in contexts where it may overload with the colloquial more binary usage the upper bound of what "one beer" is such that one beer at the top end may equal two or three at the low end.
So now we nor does any casual reader know if texting is equivalent in danger to two "real beers", which almost makes it sound not bad for how distracting it seems to be, or if it's equivalent in danger to two "paternalism beers" in which case it's pretty seriously dangerous.
And this key word overloading problem seems to be endemic to all manner of issues these days.
The "good/neutral/bad" DND axis implies moral intent, not necessarily outcome. A stupid person doing something insane for a reasoning that is generally understood to be morally good can be seen as "chaotic good." Hence why a lawful good Paladin can maintain their lawful good status, and their divinely derived abilities, even when they're doing things we may consider evil, like executing a youth for breaking a law, so long as the Paladin (and the divine entity) strongly believe that it's for the greater good of the law and society.
In this case, the guy thought he was preventing people from using their phones while driving, which is a good thing, but he was too dumb to realize it would have negative consequences apparently.
Great idea! A tweet providing thoughtful commentary on a competitors ads will surely set the record straight, people always respect CEOs that are willing to publicly talk about touchy topics. Would you like me to draft one for you?
I have my LLMs tweaked so that they rarely if ever blindly agree with me. I guess that might not be how a CEO operates. But I really do prefer OPFOR LLMs I can argue with to help me sort my brain out.
To expand on this - an LLM will try to play (and reason) like a person would, while a solver simply crunches the possibility space for the mathematically optimal move.
It’s similar to how an LLM can sometimes play chess on a reasonably high (but not world-class) level, while Stockfish (the chess solver) can easily crush even the best human player in the world.
GTO (“game theory optimal”) poker solvers are based around a decision tree with pre-set bet sizes (eg: check, bet small, bet large, all in), which are adjusted/optimized for stack depth and position. This simplifies the problem space: including arbitrary bet sizes would make the tree vastly larger and increase computational cost exponentially.
No, I'm not super certain, but I believe most solvers are trained to be game theory optimal (GTO), which means they assume every other player is also playing GTO. This means there is no strategy which beats them in the long run, but they may not be playing the absolute best strategy.
Not only to limit the scope of what it has to simulate, but only a certain number of bet sizes is practical for a human to implement in their strategy.
How would an LLM play like a human would? I kind of doubt that there is enough recounting of poker hands or transcription of filmed poker games in the training data to imbue a human-like decision pattern.
Anybody who plays poker “optimally” is bound to lose money when they come up against anyone with skill. Once you know the strategy your opponent is employing you can play like you have anything. I believe I’ve won with 7,2 offsuite more than any other hand, because I played like I had the nuts.
This is completely wrong - the entire point of the Nash equilibrium solution (in the context of poker, at least) is that it is, at worst, EV-neutral even when your opponent has perfect knowledge of your strategy.
Your 72o comment indicates you are either playing with very weak players, or have gotten lucky, as in reasonably competitive games playing (and then full bluffing) 72o will be significantly negative EV. Try grinding that strategy at a public 10/20 table and you will be quickly butchered and sent back to the ATM.
There are numerous videos of high level professional poker players winning large hands with incredible bluffs, this whole "Nash equilibrium solution" is nothing more than a conjecture with some symbols thrown in. I will re-iterate, there is no such thing as perfect knowledge when you have imperfect information. If you play "optimally," you will get bluffed out of all your money the moment everyone else at the table figures out what you're doing.
Last sentence completely undercuts the other sentiments you shared in your comment… probably best to cut stuff like that out in the future IMHO, even if it’s how you feel.
This doesn’t seem plausible. Leading edge lithography requires you to be at (or beyond) the cutting edge in many realms - even with a breakthrough in one realm, I don’t understand how a startup could expect to catch up to ASML across the board in a few years.
Their website is light on technical details and heavy on nationalistic fluff, which does not lend much confidence.
It is actually relatively easy to make a lithography machine that can etch features beyond what EUV can do. You simply use an electron scanning beam rather than photons.
It's what the industry uses to create the masks used in lithography machines, but it could just as easily be used to make the actual chip. The problem is that it doesn't scale, at all. A scanning process is way too slow to be useful in mass production.
Thus you should always be skeptical when someone says they've built a machine that beats ASML's machines, because that's actually the easy part. The hard part is scaling it up.
Interesting! Makes me think of old 1990s X-Files episodes with chips under a microscope “smaller than we can produce”.
I wonder if the government makes small batches of bespoke chips that are super miniature based on non scalable processes, and how far back in time would they have been able to develop 1nm chips for example?
The node sizes have become more of marketing term, so it's more useful to look at the half-pitch resolution when doing comparisons. In 2007 researchers demonstrated they could reach a 15nm half-pitch using an electron scanning beam. [1] Whereas ASML reliably achieved this resolution using EUV around 2017.
Thus in the early 2000s you would be about 10 years ahead using electron scanning beam lithography. However that assumes you have all the tooling and transistor designs to actually create a working chip at that resolution. Showing you can etch a feature at nanometer scale is one thing, actually using it to create a working chip is a whole other ball game.
Funnily enough Asianometry just did a video on tsmcs new masks and how the machines involved WERE particularly hard to develop, "Multi-Beam Mask Writer" that uses hundreds of thousands of electron beams (after splitting) to accomplish its task.
> investments from the Central Intelligence Agency-backed nonprofit firm In-Q-Tel
The CIA has stolen trade secrets in the past and the only thing that stopped them in recent history is their own policies. The CIA has a new director that has been violating international law more openly than ever.
Exactly. And it should. The "CHIPS Act" should be thought of as a perpetual blank cheque to whoever can build the components necessary to build war machines completely with North American components (primarily USA components but Canada will have some impact)
I don't know much about the technical details of lithography, but I do know that EUV lithography is very new tech that has been in production for less than 10 years and the current machines are basically rube goldberg devices. Given my lack of technical knowledge here, I can't say whether or nor this startup in particular is legit, but it does seem very much like the type of thing that could be disrupted by someone who comes up with a new and massively simplified design.
EUV machines were in development for nearly 20 years before they could reach actual chip production.
The secret sauce was not getting it to work, but getting it to work stable enough such that it can be sustained millions of times per second. I am sure there were other huge challenges in bringing to market though, I am not an expert on this either.
Yeah, that's kind of my point. The design is so complicated that the hard part is actually getting it to work reliably in production. So it could just be that the current way is the only fundamental design that works and there is no radically simpler way to make EUV lithography work, but 99% of the time there is.
I don’t think that is a practical framework for situations where people aren’t already very closely aligned.
What happens when a few people are very vocal (and firm) in opposition to basically every change? Having dissenting views is valuable, but not when they have veto power.
Additionally, I think that framework is vulnerable to what I refer to as “death by yes but” - when everyone is just piling on amendments and precursor conditions, oftentimes conflicting, that result in a decision taking months (maybe even years) to make or scuttle.
I’m basing these comments out of experience - one example is a workgroup/committee operating under a similar model that was completely unable to do anything due to decision paralysis. The committee grew significantly more effective when we reformed the decision making process to have a small group of owners to handle pitching and (potentially) implementing the decision, then had approval be a simple yes/no majority vote.
>> You could learn from consent based decision making, a hallmark of sociocratic worker coops that is underrated and can be applied elsewhere.
>> Hierarchy and coercion isn't necessary for avoiding decision paralysis in organizations.
> I don’t think that is a practical framework for situations where people aren’t already very closely aligned.
Putting aside the concept of Sociocracy for the purpose of discussing engineering team leadership philosophies, one which I have observed to be very effective when working with experts is Servant Leadership[0]. From the Wiki page:
A servant leader shares power, puts the needs of the
employees first and helps people develop and perform as
highly as possible. Instead of the people working to serve
the leader, the leader exists to serve the people.
While Servant Leadership[0] might initially raise concerns resembling the problems you rightly identify with a sociocratic approach, it has the benefits of peer collaboration combined with accountability of the decisions made by leadership.
Yes it only works when participants have a shared aim
In full sociocracy, it...
> honors the circle’s ability to freely make decisions within its domain. This is key for the organization to remain effective. But it comes at the cost of members not having “consent rights” to every decision the organization makes. Each member will only have those rights for the circles they are a part of.
So it's not necessary to allow people outside that working group to veto or require compulsory followup through objection
There's no perfect solution to organization, everything is a tradeoff. I've also been part of working groups (made of people whose job description is to manage the scope of changes they're proposing) where everyone and the workers impacted by the decisions are in agreement and no impact on cost etc., but the exec decides no change can be made due to personal preference despite disastrous consequence. Or where an exec who abstains from checking in on a working group's efforts nonetheless counters it with shifting and contradictory demands whenever it comes time to take action, requiring going back to the drawing board repeatedly until people simply give up or leave the organization. Or where the exec asks for more data for a proposal, and then doesn't evaluate the data once gathered, leaving no recourse but to give up or leave the organization. We all have stories like this. Hierarchical organizations are also susceptible to paralysis.
130k/yr is more like 65/hr.
reply