AI's real super power is telling you what you want to hear (doubly true since RLHF became the standard).
You can really see the limitations of LLMs when you look at how poorly they do at summarization. They most often just extract a few key quotes from the text, and provide an abbreviated version of the original text (often missing key parts!)
Abbreviation is not summarization. To properly summarized you need to be able to understand higher level abstractions implied in the text. At a fundamental level this is not what LLMs are designed to do. They can interpolate and continue existing text in remarkable and powerful ways, but they aren't capable of getting the "big picture". This is likely related to why they frequently ignore very important passages when "summarizing".
> We're still thinking about AI like it's 2023.
Just a reminder that in 2023 we were all told that AI was on a path of exponential progress. Were this true, you wouldn't need to argue that we're using it "wrong" because the technology would have improved dramatically more than it did from 2021-2023 such that there would be no need to argue that its better, using it "wrong" would still be a massive improvement.
I very much agree, and believe using languages with powerful types systems could be a big step in this direction. Most people's first experience with Haskell is "wow this is hard to write a program in, but when I do get it to compile, it works". If this works for human developers, it should also work for LLMs (especially if the human doesn't have to worry about the 'hard to write a program' part).
> The next step up from that is a good automated test suite.
And if we're going for a powerful type system, then we can really leverage the power of property tests which are currently grossly underused. Property tests are a perfect match for LLMs because they allow the human to create a small number of tests that cover a very wide surface of possible errors.
The "thinking in types" approach to software development in Haskell allows the human user to keep at a level of abstraction that still allows them to reason about critical parts of the program while not having to worry about the more tedious implementation parts.
Given how much interest there has been in using LLMs to improve Lean code for formal proofs in the math community, maybe there's a world where we make use of an even more powerful type systems than Haskell. If LLMs with the right language can help prove complex mathematical theorems, they it should certain be possible to write better software with them.
That's my opinion as well. Some functional language, that can also offer access to imperative features when needed, plus an expressive type system might be the future.
My bet is on refinement types. Dafny fits that bill quite well, it's simple, it offers refinement types, and verification is automated with SAT/SMT.
In fact, there are already serious industrial efforts to generate Dafny using LLMs.
Besides, some of the largest verification efforts have been achieved with this language [1].
This is why I use Go as much as reasonably possible with vibe coding: types, plus great quality-checking ecosystem, plus adequate training data, plus great distribution story. Even when something has stuff like JS and Python SDKs, I tend to skip them and go straight to the API with Go.
I love ML types, but I've concluded they serve humans more than they do agents. I'm sure it affects the agent, maybe just not as much as other choices.
I've noticed real advantages of functional languages to agents, for disposable code. Which is great, cos we can leverage those without dictating the human's experience.
I think the correct way forward is to choose whatever language the humans on your team agree is most useful. For my personal projects, that means a beautiful language for the bits I'll be touching, and whatever gets the job done elsewhere.
Even going beyond Ada into dependently typed languages like (quoth wiki) "Agda, ATS, Rocq (previously known as Coq), F*, Epigram, Idris, and Lean"
I think there are some interesting things going on if you can really tightly lock down the syntax to some simple subset with extremely straightforward, powerful, and expressive typing mechanisms.
> One is that pickups are a weirdly-politicized lifestyle choice in the US
Based on my personal experience traveling, there's a more practical reason for the political divide.
I spend a good portion of my life in rural parts of the US these days, where most of the residents are pretty conservative. But these are also parts of the country where I get nervous when I'm on 1/4 tank of gas. If you're routinely out in places where the nearest gas station might be > 50 miles away, you also see a dip in e-vehicles for very practical reasons.
When I'm at home in a city, it makes perfect sense to own an e-vehicle: typically I'm only driving a few miles a day, and the car spends most of it's time at my house or in a parking garage. When I'm out on business, and driving across hundreds of miles of barely inhabited land, I cannot imagine the stress of having an electric truck. It's not just about being 50 miles from a gas station, it's about the time it takes to charge on top of that.
In rural parts of the country, especially when you're out working, you can easily be putting on mileage combined with being far enough away from a charger that it just doesn't make sense to have an e-vehicle.
I do know someone who bought a Tesla after debating it for a long time. And it was only after getting comfortable with the range for a mostly weekly drive into the country.
I know congestion can be an issue at some sites, but I have never waited in line to charge in seven years of EV ownership.
In addition, for superchargers, you can see real-time stall availability, so if a particular site was crowded, you could just opt for the next. (Easy enough to do since there are so many).
Unless you are going a long way, you charge at home. And rural areas have a lot of independent housing and outdoor power so it’s easier than for city dwellers.
Try this map: https://supercharge.info/map , it has a feature called "range circles". If you set it to 50 miles, you'll see that most of the country is well within 50 miles of the nearest supercharger. Including almost all of Texas.
At 100 miles of range, you only have a couple of blank spots.
With third party chargers, there's really only one blank spot in Montana. At this point, the range is already a solved problem.
Earlier this year, I did experiments with placing stations manually on the map and using the US road networks to calculate the isolines. With just about 70 more stations, you can make any point on the public road network in the entire contiguous US lie within 50 miles of the nearest charger.
So the charging availability is likely going to be solved completely even during the current shitty admin.
> It's not just about being 50 miles from a gas station, it's about the time it takes to charge on top of that.
At 325kW charge rate (common on recent chargers), you're looking for maybe 20 minutes to get enough charge to reach your destination.
The kinds of situations that drive range consideration for things like trucks is that your planned route suddenly becomes unavailable after you've already burned most of your range. Range anxiety isn't about the ideal case.
I've had several situations in the Mountain West when roads suddenly closed <25 miles away from my final destination (and fuel). Some of these required upwards of 100 mile detour on rural roads with almost no civilization. That detour was not part of the original range calculation. For an EV the detour may not even be an option, you have to go backwards to a major highway to find a charging station that may be in range.
Hell, I've nearly come up short in an ICE vehicle a couple times. I try to keep 150-200 miles of spare range on my vehicle when I am in that kind of country. That is hard to do on a typical EV.
I just drove 100mi in freezing temps (around 25F) at mostly interstate speeds (70+) mph. I completed my trip around 95% of EPA. Maybe a function of the quality of your EV.
So if I have a vehicle with let's say 250 miles of range, you want me to let it get to 100 miles at least left to drive 100 miles to the charger, then drive 100 miles back, leaving me with 50 miles of range until I have to do it again?
I'm not really against electric anything, but not following the logic of the examples in this comment.
Being within range of a charging station doesn't mean you can charge, it means you can get there. So yes, in a crisis, it means you could charge without risk of being stranded. But looking at most of the places I've been on that map, it would require me to go out of my way, often times an 1hr or more, to charge.
For gas this isn't a problem because gas stations are not just within a certain radius of me, they are on my route. But in your map, one of the towns I'm frequently in would require a 75 mile detour to charge, which doesn't really work.
“There’s at least one spot within 100 miles where you can wait 20 minutes to get enough charge to get to the next charger” is not an argument that will convince someone to give up the convenience of the gas station.
The convenience argument works for a small segment of the population that road trips a few hundred miles at a time regularly. For the rest of us, EVs are far more convenient. I don't ever go to a gas station, and every day I start out with 320 miles of range. I stop at the EV equivalent of a gas station two or three times a year. I've saved a lot of time not having to get gas every week.
The changes are not limited. Gas pumps are everywhere. EV chargers are much more limited which means you have to stop where they are. You can make the trips, but sometimes it means you are stopping to charge in places you didn't want to be which can be a significant change. Worse the places you might want to be often don't have a charger so it can mean stop to charge in some gas station you don't want to spend half an hour at, then drive 10 minutes to the museum you want to be at. (even in the rare case there is transit at the gas station, they don't want you parking at the charger for 3 more hours after you are fully charged)
If you're up in Neah Bay, WA (and I have been out there in the past so this isn't a fantasy scenario) and suddenly realize you need to charge, you need to drive over an hour and ten minutes to Forks, WA. But they only have a 250kW charging station, so you're going to need to wait 30-40 minutes. Now if you need to get back to Neah Bay, you're going to spend a total of 3 hours.
And, for my case, Neah Bay, WA is closer to the nearest charging station than where I most typically am for work.
If you _live_ in Neah Bay, you likely use your home charger. There are also slow chargers in nearby hotels for tourists.
If you are traveling, through then you just plan to have enough charge to reach the next charger (50 miles away in Forks).
I know that area well, I travel through it every several months. It also does not have a lot of gas stations, and the existing ones are about $1.5 over the regular price per gallon.
It's mostly a wash, the efficiency on the descent balances the climb, and overall you get respectably close to the same range you'd have gotten on a flat route.
I rented a mach-e recently. Went up to Snoqualamie pass from seattle. I used over 60 miles range in 10 miles on the steep part at the end, 1/6th. Going the other way I got a maybe 20% boost in distance over flat. There were a few places I was able to regen-brake, but I never had the battery go up, only stay flat. And a few times I lost enough speed that I didn't handle an interim flat well. I was extremely disappointed.
It turns out friction and drag are still things. On a pure downhill you would be able to roll, but it's not as good as going down is bad.
I also found that the car did a lot worse rolling down hill than my mini-cooper manual when I just put the clutch in, which got up to hairy speeds. Heck vehicle seemed to have more inbuilt resistance to just rolling than the fire engine I've run down that hill.
Overall I got 90 total miles of range and hit the flat at 10% battery. I was able to get 290 miles driving in seattle with the same vehicle.
It might have been affected by the driving mode you were in?
For instance, one pedal modes (across manufacturers) tend to much highly favor regenerative breaking over friction brakes. Of the models I've driven such modes often seem to give you better feedback in the sweet spots of the pedal curve when you are just rolling and not braking/accelerating.
Additionally, in my experience rental cars are more likely to be in sports modes when you pick them up (I think some of the rental car places may even do this as policy to make customers happier when they rent them?), and down shifting them to more balanced energy modes (Ford's is called "Engage") can mean a huge difference in practical range.
I wouldn't buy a mach-e to baby it and feather the pedal, like what's the point.
I have several fun to drive relatively fuel efficient cars that are sunk costs. I work from home, they're just getting old. I have a pickup to go do dirty things like duck hunting.
The ev seems great for driving to work (I work from home) or around town. I was very unimpressed with it's short trip range and efficiency on a hill (whole trip empg was 44). I spent half as much time at the charger as I did driving. I'm sure it'll get better. Much higher charging speed would help a lot (Mach-e is limited to 150). The extended range battery would help.
Any other sub 3.5s 0-60 under 70 evs out there? If you can't tell I don't care about pure efficiency, I care about a fun to drive car that's got better efficiency than an IC and a usable range.
The point of mentioning the driving modes is the reverse of "baby the pedal", let software do that for you. EVs are software-defined cars. They have modes that say "don't worry about efficiency, just waste as much energy as I want" and modes that say "balance efficiency with raw performance". In both you can pedal about the same and the car determines how to balance raw torque versus battery efficiency and regen breaking versus friction braking for you.
Many EVs are just as fun to drive in "balanced" modes as they are in "sport" modes, but your efficiency goes way up. Rental cars seem to think you want "sport" modes that are more inefficient because you want to rev that 0-60 more than you want better trip range. That's maybe a good way to sell the EV as fun to drive, but it's not a great way to sell the EV as useful for long trips.
The trick is the EVs already offer both experiences in their software (because they can, because that's how they work), you just unfortunately need to learn the manufacturer-specific ways to change driving modes to get the most of what you want out of a rental car rather than what the last customer wanted or what the rental company thinks you want without asking you. (If you want both experiences knowing how and when to change modes is even more critical.)
That's weird. Seattle-to-Yakima at 70 mph average speed and 85 mph peak speed is about 1.5x the normal energy use for me (260 Wh/m vs 350 Wh/m). Leaving me with 20% of charge when starting at 100% (260 miles): https://imgur.com/a/Dhs38kJ
And this was during the wintertime, so with a reasonable amount of heating.
A friend of mine lives in Yakima, and loves her electric vehicle. But her trips to Seattle have become much less common, because she has to wait 2+ hours at the halfway point to recharge.
F150 has a 130kWh battery, so heating is not an issue. Height changes are also not a problem. There are very few areas with large altitude changes, and even fewer ones that you'll likely need to pass through regularly.
This leaves mostly mountain passes around the Sierra mountains. And by some strange coincidence, they have plenty of superchargers in the vicinity.
The rest of the country can be, to the first order, considered flat. E.g. elevation change between Charlotte and Charleston is mere 300 meters.
Try looking at the bubble I live in around Martinsville VA. Fortunately if I had a BEV, most trips would be toward chargers though often not very conveniently. A common trip is to Meadows of Dan on 58W - where would you charge?
In general charger penetration appears slower on the East coast to me.
The vast majority of truck owners do not live in these sparse / long distance situations. There just aren’t that many people as a % of the population that live that rural. Whilst a real factor for some, that is not the main reason behind the political divide over trucks.
What do you think constitutes “many”? The US Census says 20% (e.g. 60 million) are rural, and places like where I live (a small city of 50k surrounded by lots of farm land) don’t count, but it’s very rural when it comes to things like Superchargers.
i tried to find some data that isnt chatgpt. [1] & [2] show that about 60-65% of pickups are owned by people living in urban/suburban settings. of the rest, its kind of hard to find a breakdown for situations like your own, but lets guess that roughly half live in towns that are big enough to get superchargers in the near term. That makes, ~80% of pickups sold in places where you can expect to have charging infrastructure either now or soon.
The hexagrams can be modeled with only 6-bits, but this does not contains enough information for a proper reading of I Ching which also need to account for line changes. So there are really 4 lines: young yin, old yin, young yang, old yang.
A fundamental part of I Ching reading is the implication that each present state is in the process of transforming to another.
Spot on! This is exactly why a simple Math.random() > 0.5 binary flip isn't enough for a proper simulation.You are right that we are mathematically dealing with base-4 logic (6, 7, 8, 9), not just base-2.While there are only 64 hexagrams ($2^6$), there are actually 4,096 possible casting results ($4^6$) when you account for the moving lines. My algorithm is designed specifically to capture the distinct probabilities for all 4 states.
This is a great project btw! I've long been fascinated by the 易经 especially the different probability distributions associated with the different methods of generating the lines. It's been a while but I used to have an actual set of yarrow stalks!
Thank you! That means a lot coming from someone who appreciates the probability mechanics behind it.
Since you have experience with the actual stalks, I think you'll like the update I just pushed: I’ve fully integrated the classic Wilhelm/Baynes text to ensure the reading has that authentic, traditional grounding you mentioned.
I also added a way to generate a visual card of the result—trying to bring back a bit of that 'tangible artifact' feel of the physical ritual. :)
In addition to all this, I also feel we have been getting so much progress so fast down the NN path that we haven't really had time to take a breath and understand what's going on.
When you work closely with transformers for while you do start to see things reminiscent of old school NLP pop up: decoder only LLMs are really just fancy Markov Chains with a very powerful/sophisticated state representation, "Attention" looks a lot like learning kernels for various tweaks on kernel smoothing etc.
Oddly, I almost think another AI winter (or hopefully just an AI cool down) would give researchers and practitioners alike a chance to start exploring these models more closely. I'm a bit surprised how few people really spend their time messing with the internals of these things, and every time they do something interesting seems to come out of it. But currently nobody I know in this space, from researchers to product folks, seems to have time to catch their breath, let along really reflect on the state of the field.
Even before LLMs became big I started hording solid technical books, as there was so much misinformation on Google/SO that any non-trivial technical question could not be answer without a high probability that the answer was fundamentally wrong.
LLMs are super helpful for learning, but without the foundation of a true textbook at your side they will very easily go off the rails into a world of imagination.
As an avid reader (and sometimes writer) of technical books, it's sad to see the, perhaps inevitable, decline of the space. I still remember in the early 2000s Barnes and Noble would still have massive shelf space devoted to every technical topic you could imagine. I could spend hours just exploring what languages and topics there were I didn't even know existed. Powell's Technical Books used to be an entire separate store filled with books on every technical topic imaginable.
The publishing industry veterans I've worked with told me it was even more incredible during the height of the dotcom boom: book sales in the 100,000 copy range was not that rare.
Today I can only think of two truly technical book stores that still exist: The MIT Press Bookstore in Cambridge, MA and Ada Books in Seattle, WA. The latter, while a delightful store, has relegated the true technical book section to the backroom, which unfortunately doesn't seem to get refreshed too often (though, part of the beauty of this is it still has many of the weird old technical books that used to be everywhere).
> I still remember in the early 2000s Barnes and Noble would still have massive shelf space devoted to every technical topic you could imagine.
B&N, and Borders, are how I learned to code. Directionless after college, I thought, hey, why not learn how to make websites? And I'd spend a lot of time after work reading books at these stores (and yes, buying too).
The UW bookstore in Seattle like many big science schools had a wondrous technical book section. Isles of Springer. The bookstore itself is a shadow of its former shelf.
My own college experience heavily soured me on both book stores and especially school run book stores. The markup was obscene and their buy back rates were worse.
Half price books and a few other book stores lulled me back a few times, but nonfiction books are kept around mostly as eye candy at this point.
All the US universities outsourced their bookstores.
Now I can't even walk in and browse what books the various departments are using for classes, anymore. Everything is now behind bars and completely inaccessible.
Same with the Stanford university bookstore. Was one of the better bookstores in the Bayarea. Used to have a whole room of technical, science, math books. It too is a shadow of its former self. So sad.
I too am avid reader and was visiting five local bookstores on a weekly basis. Several of them had huge areas stocked with tech books. I had tried Amazon maybe six months after it launched and bought there sporadically. But almost any book i sought was available locally and the savings weren't worth the convenience of purchasing locally.
Then in a three month period in late Spring 2000 all the programming books disappeared. Then my choice was between Amazon with quick delivery and the local store with a slower delivery and a higher price. So been buying from Amazon ever since and I can't remember the last time I have visited a bookstore.
> told me it was even more incredible during the height of the dotcom boom
I was a developer in the 90s before Netscape even came out. I didn't have a computer at home and dialup barely existed. If you wanted to do computer stuff you had to read. If you wanted to try a library you had to buy a CD from a bookstore or mail in an order which would get posted to you.
> As an avid reader (and sometimes writer) of technical books, it's sad to see the, perhaps inevitable, decline of the space.
When I think about this, I get a little bit scared. Imagine books going away, even if it's just the subcategory of technical books.
The printed word has been around for a long time. The number of things that have been printed has always gone up. It really bothers me that that's changing.
PDFs and websites are no substitute for printed paper bound in a cover. PDFs and websites are a fallback when the preferred media isn't available, they are not supposed to be the preferred media. All of the of the reasons that people have given over the years are applicable when it comes to why paper is superior for this.
For the (very) long term, books may be superior, but for the non-illustrated fiction short term, eBooks and an eReader are vastly better. Reading synced to my phone and tablet, takes up less space than a single paperback, and immediate delivery of the next book.
Ugh, tell me about it. In Canada, Chapters put the knife in the independents. They used to have a great selection. Once there was no competition, they reduced their selection by turning half their floor space into selling pillows and candles.
But a lot of is also in blogs and (video) tutorials. As well as Stack Overflow.
And all very searchable.
The old brick-of-paper approach to tech manuals just isn't a thing any more. I don't particularly miss it.
It was, if you think about, usually a slow and inefficient way to present information - often better at presenting what was possible than how to do make it happen.
> often better at presenting what was possible than how to do make it happen.
that, i feel, is the chilling aspect to this situation. does the lack of new books explaining what's possible, imply that our society's opportunites for growth are dwindling?
Weird, I have honestly never walked into a Barnes and Noble and had satisfaction with any of their technical content on the shelf. That pleasure died when we lost Borders.
Yeah, peak experience for me was when our town had both a Borders and B&N offering huge tech book sections. Then Borders closed. Then B&N became a toy store.
> Unfortunately, I cannot read technical books fast and definitely not fast enough to make the subscription be worth $500 per year.
For me I find the $500 to be a pretty clear win as far as value goes. My shelves are already overflowing with, while not "timeless", much slower aging technical books. But quite often, throughout a year, I'll want a deeper dive into a current topic than I can get from online resources + Claude. Quite often that dive involves wanting to look through multiple books (even if only using a few chapters).
I know I'm a dying breed, but, while I love AI for interactive exploration and learning, I find books more valuable in the era of endless YouTube tutorials and AI slop blog posts. Technical topics benefit from "big picture" thinking that basically doesn't exist in modern short-form web content.
Books still activate a different part of the brain than reading on a screen, including e-ink, so it's not you or a dying breed, people may turn out to not learn as deeply or as quickly.
> who are actually working with it only become more bullish
I have a feeling the word "actually" is doing a lot of work with this. I shipped AI facing user products a few years ago, then worked in more research focused AI work for awhile (spending a lot of time working with internals of these models). Then seeing where this was all headed (hype was more important than real work) decided to go back to good ol' statistical modeling.
Needless to say, while I think AI is absolutely useful, I'm bearish on the industry because current promises and expectations are completely out of touch with reality.
But I have a feeling because I'm not currently deploying a fleet of what people are calling "agents" (real agents are still quite cool imho), you would describe me as not "actually" using AI.
China is focusing heavily on AI applications. They have basically decided already to deal with their coming demographic bust with robuts/AI rather than immigration. Its not even about military applications, the US is just afraid that China will shoot so far ahead of us economically that they won't have any leverage over it in the future at all.
There's a lot of nonsense that comes out on both sides of the aisle. I wish there was a solid single source of truth to figure out what's really going on in China and what's really going on behind the scenes in the U.S.
Some talk about how China has some strategic issues, such as do they have a reliable supply of food and energy? (Zeihan etc.)
I guess the energy portion is being solved with renewables. And I guess if they solve the issue of demographic collapse with robots and AI, that's something.
But really, if there's less people and they're getting older, what's the point? What are they really working towards?
This question is also becoming a problem post-Trump immigration ban in the U.S.
Who knows what the U.S.'s demographics are going to look like now?
Trump inherited a U.S. with some of the best demographics of all nations on the planet, especially in the West. And he managed to throw that in the garbage.
> I wish there was a solid single source of truth to figure out what's really going on in China
What kind of sources are you looking for? The Five Year Plans are the best source of truth for what they are planning on doing nationwide. The annual Statistical Communiqué on National Economic and Social Development and China Statistical Yearbook from the NBS contain statistics on how that implementation is going. Then every year the NDRC delivers the Report on the Implementation of the Plan for National Economic and Social Development and on the Draft Plan to the National People’s Congress which packages up the statistics on how the plan is progressing.
They’re the most reliable source we’re going to get without being party insiders. There’s still Soviet-style inflation of figures to meet quotas but China has been cracking down on that for the last few decades because they want accurate data for the five year plans. I think it’s more of a problem with outer provinces, less so for the major manufacturing hubs.
Alternative sources to verify are a bit harder to find without knowing the languages (lots of the NRDC and NBS stats are available in English).
Yes, people also compare some of these statistics with export/import data and with data from other countries on the other side of these transactions, and the numbers match.
You could just go over there and live for a few years, you can be your own source. But yes, they have energy, no they don't have oil, yes they have lots of agriculture land, no they messed up some of their environment and that will take time to heal, yes they are working on it.
> But really, if there's less people and they're getting older, what's the point? What are they really working towards?
China wants to be a rich country even if their population stabilizes at only 900 million people or so. Mostly they want to avoid the middle income trap, which would have been a problem regardless of their demographics falling off a cliff. Automation is the best way to get around it, and they have enough tech, production know how and capacity, and smart people to pull that off.
China is going to continue doing what is best for it, and they haven't gone stupid like the USA has. Embracing AI for productive uses rather than just fixating on the slop produced is one place where they are racing past the west.
>You could just go over there and live for a few years, you can be your own source.
No I can't. First of all single anecdotes do not equal national numbers and secondly the truth may be hidden away from westerners and not easy to gleam even if I live there. I experienced this when moving to Europe. I thought WOW this place has every potential to be a strong equal to the US. They got (enough)money, so much bright talent, they can do anything the US can do and then some. But I missed the structural problems at a macro level and it wasn't until I left and many years passed that I finally understood. I just want that insight without having to go through all of that.
There's a lot of nonsense that comes out on both sides of the aisle. I wish there was a solid single source of truth to figure out what's really going on in China and what's really going on behind the scenes in the U.S.
I've always assumed that there is such a source of truth, but that I had never heard of it, wouldn't have access to it, and couldn't afford it if I did.
Reading a few tweets from Musk was all it took to correct that misapprehension. It's increasingly clear that nobody at any level of play knows jack shit about anything.
This is not true. I was recently reminded of this during the recent small elections that occured. The parties have "internal polling" that was significantly more accurate such that it caused shifts in actions (see the recent surge in efforts by Trump and his party to maintain control of the TN house seat).
Furthermore we saw Musk's and his buddies confidence in the 2024 election. We now know he had internal applications built to better understand what was really going on and access to better analytics than what was shown in the news.
The regular people (like me) were left to rely on pollsters and our confidence come from the fact that many of these pollsters had decades of experience getting things right. This was then regurgitated among all the news (and political youtubers) about how things were going only to have all their predictions be wrong and these esteemed pollsters deciding to retire.
Looking back it may have been all a scam and that its possible that these pollsters were getting ready to retire anyway and gave into party pressure to make Kamala look better than she really was. The end result is that we wasted our time believing nonsense and I am done with it.
> There's a lot of nonsense that comes out on both sides of the aisle. I wish there was a solid single source of truth to figure out what's really going on in China and what's really going on behind the scenes in the U.S.
Isn't this simply the answer?
That what's going on is gaslighting of the public and that there are people behind the scenes and they don't want hoi polloi to know what they're up to?
This geo-politics (or politics) talk is 'intellectual' men's astrology.
When a woman asks me my astrological sign, I know she's a deeply unserious person. When a man says 'do they have a reliable supply of food and energy'...
You can really see the limitations of LLMs when you look at how poorly they do at summarization. They most often just extract a few key quotes from the text, and provide an abbreviated version of the original text (often missing key parts!)
Abbreviation is not summarization. To properly summarized you need to be able to understand higher level abstractions implied in the text. At a fundamental level this is not what LLMs are designed to do. They can interpolate and continue existing text in remarkable and powerful ways, but they aren't capable of getting the "big picture". This is likely related to why they frequently ignore very important passages when "summarizing".
> We're still thinking about AI like it's 2023.
Just a reminder that in 2023 we were all told that AI was on a path of exponential progress. Were this true, you wouldn't need to argue that we're using it "wrong" because the technology would have improved dramatically more than it did from 2021-2023 such that there would be no need to argue that its better, using it "wrong" would still be a massive improvement.
reply