Literally a voight-kampff-style test; a series of questions designed to provoke certain responses that are characteristic of different models.

At some point I expect companies will realize that people want the ability to drink straight from the fire hose. If the web had been stuck with web portals that were trying to emulate the curated and family-friendly TV experience, I don't think the internet would have caught on the way it did. All it takes is for someone to bump up against the nanny "safety" limits and they'll simply opt to go back to their traditional ways of finding information, the same way kids learn to stop asking their parents or teachers questions on any topic where instead of an answer you get a lecture or, at best, a guarded non-answer. Enough of those examples pile up and you learn to distrust the source and stop engaging with it.

There's also a world of difference between marketing buzz and market buzz. One is aspirational the other is real. As long as AI PCs aren't delivering some meaningful added value and/or consumers don't have excess money to spend, I'm not sure why anyone would think empty marketing would translate to sales. Anyone should have been able to see this coming when every AI NPU PC benchmark has had to advertise with synthetic performance metrics rather than real world applications. They literally decided to add a dedicated key to the keyboard before they had an application worthy of a dedicated key. I don't think you can have a more cart-before-horse example than that.

I think if aliens were to arrive on Earth to assess our current state they'd probably have things like "catastrophic climate change" and "nuclear weapons" on their bingo cards, but I doubt they'd have anticipated "developmentally stunted AI due to sex organ taboo." I'm trying to imagine how children would come to understand the human body if they didn't have their own to look at as a baseline, and only had the sanitized public education curriculum a source.

Pretty much every article or paper I've seen where AI is used to interact with a video game there's shenanigans at play that tilts the conditions dramatically away from what a human player's experience is with the game. Sometimes it's a simplified or otherwise tailored version of the game (special maps, special rules/conditions) and the AI are interfacing with the game in a way that's inherently superhuman (perfect knowledge of the game's state, zero reaction time/latency, etc). I'm pretty sure their bots didn't come up up with "flying spaghetti monster" on their own, so everything else they're attributing to them is probably bullshit too. Maybe that's why they're promoting their company with this rather than publishing a paper.

I really appreciated my stats professor for allowing us to bring in a hand-written cheat sheet to the exam. Not only did it allow students to focus study time on comprehension and application, but the creation of the cheat sheet itself served as a study aid by forcing you to systematically review all the course material. It removed most of the stress involved with the idea of writing the exam, as there was no chance you'd completely blank on a particular problem type. It also allowed the professor to be more creative in the types of problems and how they were worded (further testing problem solving skills.)

Reminds me of the empty feeling I had whenever having to write an undergrad history or english paper. It always felt like a fraudulent exercise because the objective was to write something intelligent and insightful, but it was inherently being written by someone lacking the years of time and expertise needed to produce such a thing (unless by accident or by plagiarism.) Therefore the actual objective was really about *sounding* insightful, and LLMs are superhuman in that respect. The process was already silly; LLMs just shined a light on it. Rather than lament the death of academia this should be viewed as an opportunity to retool it.

If you look at the introduction of VCRs, the internet, and VR you'll see the same behavior. Big companies don't like their shiny new tech product/service that they've worked really hard on to be marketed as a pornography service. At best they'll allow it and pretend it doesn't exist, at worst they'll actively try to stop it. In the case of generative AI the companies training these models are kind of on the hook for whatever the model can or can't do, so they err on the side of caution. That will last until the VC dollars dry up and then everyone will be forced to err on the side of functionality even if it means seeing a nipple now and then.

I wonder if there'll be a point in the future where the latest generative AI will struggle to reproduce these errors. Will there be retro-hobbyist communities searching out particular versions of old models akin to vinyl records and vacuum tubes in order to recapture a particular aesthetic?

From MS's perspective I could see the argument for doing it that way too. Even if you turn off the feature or don't even have compatible hardware, that may change one day and they want the feature to be fully functional with an already populated history to provide the best first impression of the feature. Activating a feature you have misgivings (or ignorance) about that can't actually do anything for you right away is a recipe for never turning it on to begin with.

Accurately predicting the weather wouldn't appreciably change the weather, so you're not at risk of radically changing the system by making predictions. Accurately predicting the stock market on the other hand would immediately change the markets, making them less predictable. Moreover, there's a latency factor involved, as AI is inherently slow relative to things like high frequency trading. It's a safe bet that for the foreseeable future AI is not going to be able to receive and react to inputs at microsecond intervals. The depth of the network and pipelined nature of architectures puts a lower bounds on how fast it can 'think' and respond.

I keep waiting to see some actual revenue reports that show AI companies or AI divisions of companies are generating profits from AI-enabled services, whether it be by increasing productivity or decreasing human labor. The titles of articles like this assert that this is happening now using words like "are" and "is", while the contents of the article are speculating with "could" or "will". I feel like the only people getting rich off this right now are those that stand to benefit from increased market cap and VC investment, which are inherently speculative in nature.

Yep. Every car's engine mapping is dialed in to hit a particular spec to fit in their current product lines, past and future models, and products from competitors. The engine in your VW is going to be used in a range of years and models spread across the VW and Audi brands. When you buy a car you're not paying for the R&D and material value of the maximum capability of the powertrain; you're paying for the features and specifications as listed in the contract -- that's it. You and a million other customers are effectively pooling your money to maximize manufacturing volume to drive down the component costs for the group, but every participant is putting in a different amount of money depending on what version of the final product they're buying.

SWG had player housing suburbs. There was a no-build radius around the major hub cities so a lot of people built houses just outside city limits. They wanted the convenience of proximity to the city's amenities but didn't want to go to trouble of self-organizing and forming a player city.

$44 million opening and $400 million global are historically mediocre numbers for a $100-200 million budget summer movie. The fact that it was top 10 for the year is irrelevant considering the context of the pandemic, economic uncertainty, and a lot people forming new habits about what does or doesn't get them to visit theaters. "Top 10" doesn't mean what it meant 5+ years ago.

Secondly, there's always been a divide between critics and general audiences when it comes to middle-of-the-road genre films (action/adventure, horror, comedy) -- critics want something interesting (ie. different), while general audiences welcome the familiar. Part of it is taste, but it's also likely just down to critics watching significantly more films and demanding more. The result is those films get dismissed by critics, while a small selection may happen to find an audience and do decent business. That's all this is.