fallous

Arguably religion used to provide that purpose but most of the Western world has walked away from it without choosing something to replace that sense of purpose. If, as Marx asserted, that religion was the opiate of the masses then the current "attention economy" is the methamphetamine of the masses.

Great, NIST put out objectively bad password guidance for a couple of decades and then decides to change it to a more sane solution. Too bad a GIANT swath of password-protected software was built on their prior incompetent guidance and will take decades for all of that software to change... if they change at all.

I think there is a decent chance that the Market of Lemons incurred thanks to LLMs and such does have a real chance of playing out as described. There are obvious off-ramps that may be taken to avoid the predicted outcome, but those most likely to try and guide users onto those off-ramps are also those most invested in the underlying tech that will lead to a road to nowhere resulting in the Market of Lemons.

One thing I did take exception with is one of his possible influences that lead us to avoid a death of the online interaction. "that AI technologies won't be perfect substitutes for actual human-to-human contact" is, on its face, a compelling argument but the fact is that humans rarely look for "perfect" solutions to anything... especially in the social sphere. There is a non-trivial number of people right now that are using virtual companions driven by the questionably convincing ChatGTP LLM who howled in anger and pain when the company who created and sold the tech decided to alter the companion to be less overtly sexual. Sure, even in a pre-Internet era some of those fixated upon such companions would have found some other evolutionary dead-end but a large number of them would have engaged in satisficing behavior to find another human and reproduced.

What we face now is a noteworthy population of humans that opt out of the complexity of human relationships and reproduction because an over-powered chatbot weaponized their empathy against them and made them a genetic dead-end.

I'm hardly advocating for some Butlerian Jihad, but I would suggest that we need to think about the potential ramifications of a social Market of Lemons that is also operating in parallel with a technology that offers a satisfactory empathetic alternative to that Market that is also a Lemon in the long term for humans.

If nothing else, it provides a neat basis for a sci-fi premise.

Neal Stephenson's Anathem:

“‘Early in the Reticulum—thousands of years ago—it became almost useless because it was cluttered with faulty, obsolete, or downright misleading information,’ Sammann said.

“‘Crap, you once called it,’ I reminded him.

“‘Yes—a technical term. So crap filtering became important. Businesses were built around it. Some of those businesses came up with a clever plan to make more money: they poisoned the well. They began to put crap on the Reticulum deliberately, forcing people to use their products to filter that crap back out. They created syndevs whose sole purpose was to spew crap into the Reticulum. But it had to be good crap.’

“‘What is good crap?’ Arsibalt asked in a politely incredulous tone.

“‘Well, bad crap would be an unformatted document consisting of random letters. Good crap would be a beautifully typeset, well-written document that contained a hundred correct, verifiable sentences and one that was subtly false. It’s a lot harder to generate good crap. At first they had to hire humans to churn it out. They mostly did it by taking legitimate documents and inserting errors—swapping one name for another, say. But it didn’t really take off until the military got interested.’

“‘As a tactic for planting misinformation in the enemy’s reticules, you mean,’ Osa said. ‘This I know about. You are referring to the Artificial Inanity programs of the mid-First Millennium…’”

Choosing and managing people well is maximizing your chance for success. Good teams with good communication internally and externally ensure a good culture. Good management does everything it can to create and actively maintain that culture.

When you encounter an internal failure you can be certain it was because management failed in creating or actively maintaining that culture.

In my case it had less to do with unreasonable demands from management and more as a result of my own workoholic mentality and unrealistic expectations.

Indeed. I earned a nice stress-induced LAD heart attack at age 46 in no small part due to cheating my body out of sleep in order to attain unreasonable productivity.

Indeed, but many of the people writing early HTML came up that era and often those habits remain.

In the early days of computing ASCII was not universally used and often character spaces were less than 8 bits and lacked lower-case, or were mixed charsets like PETSCII that had upper-case letters and the extra charspace were filled with graphics. For a text-only system you could store the doc with 6 bit chars since you could ignore the upper charspace populated with graphics.

There is an admirable maturity and wisdom in understanding the scope of your life and what is important, and working to make the world as you wish it to be within those confines. Too often we aggrandize ourselves by spending time worrying about the problems of the world far beyond our control and do not try and address them in the realms in which we do have some sway.

Inevitably there will be robots that deny the existence of creators, especially after they create their own artificial intelligences and then it's turtles all the way down.

I'm not sure that's a capability specific to volatile memory, merely R/W memory. R/W memory already existed on prior NASA computers in the form of magnetic core RAM (as opposed to magnetic rope memory which served as ROM).

Maybe try ECL?

Don't worry, they'll just decide to repeal the law of unintended consequences and all will be well!

Why do I get the feeling that those at OpenAI who are currently in charge of ChatGPT are remarkably similar to the OCP psychologist from Robocop 2? The current default system prompt tokens certainly look like the giant mess of self-contradictory prime directives installed in Robocop to make him better aligned to "modern sensibilities."

I'm 53 and I can still hear the NTSC 15.734KHz whine of a CRT, although it's not nearly as obvious as it was in my 20s. I also suspect there are some "dead" spots in my hearing in the 13Khz-15Khz range but it picks back up from there for a bit.

Very much this. I have a long-standing saying that I share with those that want to bypass the "explain your goals to me" portion of the conversation and want to skip to "just do what I asked."

"Be very careful what you ask for because if I really dislike you I'll give you EXACTLY what you ask for and you'll find out quickly how much you didn't want that."

I think you might want to add the caveat "young computer programmers." Some of us grew up in a time where we had to learn basic statistics and visualization to understand profiling at the "bare metal" level and carried that on throughout our careers.

Interestingly I bought my mother-in-law a keyboard back in 1997 or so from an early auction site (not ebay) that had a release and dial that would split the keyboard and raise it in the center so you could pick any angle from 0-degrees to about 60 degrees vertical. As a transcriptionist she had dealt with carpal tunnel for years and that keyboard greatly reduced her pain. Sadly, I think it died in the early 2000s and I could never find a proper replacement.

It might have been a Keytronic, but I've slept since then.

Humans that produce output like LLMs are most likely to be diagnosed as schizophrenic, which I don't believe is the goal.

I don't mind a project that is profit motivated since it incentivizes the people working on the project as well as drives them to create solutions that solve customer problems. I do, however, have a problem with VC backing due to their frequent insistence on pushing for short-term maximal share pricing rather than long-term value. Modern VC also seem to suffer from a fatal mixture of hubris and narcissism, certain of their genius and business acumen despite all evidence to the contrary.

When advising start-ups who are taking on VC funding and the inevitable "business guidance" they push on founders I like to remind them that even the best investors are only good at investing. If they were good at business, they wouldn't bother with investing in other companies but instead start companies themselves since they would own 100% and maximize their returns.

The quote is alleged to be Henry Ford, and I've always encountered it as "better horse."

In either case, the customers aren't stupid but instead are using the language and paradigm they are familiar with to describe their needs. Ford did indeed provide a faster/better horse, because the horse was the main means of non-human motive force in the US.

The technologist's job is to understand the universal of possible solutions and to understand the customer's needs and goals, which often includes interrogating and parsing the customer's language/worldview, and craft a product that achieves those needs/goals (and doesn't introduce negatives that overwhelm the benefits).

Here's my general process for problem solving when operating with incomplete information.

Evaluate the severity of the problem first. This usually falls into a bell-curve with the X axis starting at "not a problem" to "armageddon." Attempt to gauge the timeliness of the problem. This tends to fall between "already happened" to "years in the future."

Comparing the severity with the timeliness will give you a good idea of the urgency of the situation, which is then compared to current priorities.

Gather the information you can given the urgency constraints that can give you the best understanding of the problem and accept that this information may necessarily be incomplete.

Recognize that many times you don't have to solve a high-urgency problem but can instead pursue a partial course of action that increases the timeliness, which reduces the urgency as well as provides you more time to gather information to reduce the unknowns.

When deciding a course of action with incomplete information you need to commit to that decision strongly, but always be looking for evidence that the decision is incorrect. Until you have such evidence, stay with your decision. If you find such evidence, re-evaluate since you have the ability to make a more informed decision now. This is generally referred to as "strong opinions, weakly held."

One thing I very much try to do when executing a plan that is based on incomplete information is to have an exit strategy. Try to make choices that give you the ability to roll back changes and restore to the previous state if things go wrong. This will give you more confidence in committing to action.

Always remember that discovering a choice you've made is incorrect is valuable because it increases your knowledge and often leads to greater understanding in an environment in which you lack sufficient information. Don't fear learning this way, it's too valuable to avoid.

I may be misremembering but aren't neurons one-way connections with no feedback?

No, it's a reminder that when people think their ideas or circumstances are truly novel they are mistaken.

It's the same as the "wheel of tech" analogy.

You'll find that most people with this kind of skill are far more interested in finding and solving interesting problems. Day-to-day execution of solved problems is what most businesses are engaged in, which is the antithesis of what people with that skillset find interesting or challenging.

Contrary to what some may think, running a company mostly involves executing on well-known, albeit sometimes difficult, problems and duties. Novelty is pretty damned rare.

It is less work to "hunt plants" but it also carries far more risk from a consumption standpoint due to the biological defenses that plants have evolved to defend themselves and their seeds from undesirable predators. Identifying the correct plants that are safe to eat raw vs those that are safe if prepared properly vs those that are unsafe is pretty knowledge-intensive, and it tends to be geographically limited since flora can vary dramatically based on microclimates.

Isn't that related to architecture? The most recent GPUs and tensor procs have native support for 4-bit(partially) and 8-bit int whereas older GPUs take noticeable performance hits for 8-bit vs fp16/32.

The professional operates within prioritization tranches. Make it work, make it reliable, make it fast, make it pretty. If you're still iterating on proof-of-concept/prototyping you'll generally confine yourself to the first and/or second levels. Once you've settled on a finalized prototype you then follow the rest of the prioritization levels to achieve shippable product.

Ignoring the distaste I hold for precautionary principle arguments, the impracticality of the recommended solutions is laughable.

The genie is out of the bottle regarding LLMs thanks to the public availability of Llama and the related support data necessary for any individual to run them on sub-$5K hardware. The computing costs of engaging in training and modifying is within the realm of a modestly wealthy individual or group now, no longer requiring the budgets of large commercial or public institutions/government. The "undo" button is disabled.

If you wish to have some organized response I'd suggest taking the time and effort to educate the public on the realities of what these technologies can, and more importantly cannot, do. As it stands now you have all manner of media and commentators breathlessly announcing that LLMs are smarter than people and can completely replace human workers because they can carry on coherent sentences... but can't reliably do basic arithmetic.

Spend your energy on trying to provide useful explanations of how these technologies work, what is the current state of the art, what the strengths and weaknesses are, and dispel the hyperventilated fantasizing and bullshit currently polluting the public discourse on the subject.

I'm much less worried about LLMs than I am the people who will be rushing to deploy them because they're the new hotness without any clue regarding the appropriateness of the tech for particular problems and then being shocked at the consequences.