Computer scientists are, in part, responsible for the public confusion about what LLMs are and can do. Tech investors and founders, however, are the biggest liars and BS peddlers when they go around saying irresponsible things like LLMs are on the verge of becoming "conscious" and other unfounded and impossible things (for LLMs). It's not a surprise that many people believe that you can have a personal "conversation" with a tool that generates text based on statistical analysis of previous data.
1. Ars Technica's (OP website) audience includes tech enthusiast people who don't necessarily have a mental model of LLMs, instruction tuning or RLHF.
2. why would this "study" exist? - for the reason computer science academics conduct study on whether LLMs are empirically helpful in software engineering. (The therapy industrial complex would also have some reasons to sponsor this kind of a research, unlike SWE productivity studies where the incentive is usually the opposite.)
For the record, my initial question was more rhetorical in nature, but I am glad you took the time to share your thoughts as it gave me (and hopefully others) perspectives to think about.
> It's a bit like talking about the quality of pastoral care you get at Church.
No, no it isn't.
Whatever you think about the role of pastor (or any other therapy-related profession), they are humans which possess intrinsic aptitudes a statistic text (token) generator simply does not have.
A human may also possess malevolent tendencies that a silicon intelligence lacks. The question is not if they are equals, the question is if their differences matter to the endeavour of therapy. Maybe a human's superior hand-eye coordination matters, maybe it doesn't. Maybe a silicon agent's superior memory recall matters, maybe it doesn't. And so on.
> A human may also possess malevolent tendencies that a silicon intelligence lacks.
And an LLM may be trained on malevolent data of which a human is unaware.
> The question is not if they are equals, the question is if their differences matter to the endeavour of therapy.
I did not pose the question of equality and apologize if the following was ambiguous in any way:
... they are humans which possess intrinsic aptitudes
a statistic text (token) generator simply does not have.
Let me now clarify - "silicon" does not have capabilities humans have relevant to successfully performing therapy. Specifically, LLM's are not an equal to human therapists excluding the pathological cases identified above.
> Let me now clarify - "silicon" does not have capabilities humans have relevant to successfully performing therapy.
I think you're wrong, but that isn't really my point. A well-trained LLM that lacks any malevolent data, may well be better than a human psychopath who happens to have therapy credentials. And it may also be better than nothing at all for someone who is unable to reach a human therapist for one reason or another.
For today, I'll agree with you, that the best human therapists that exist today, are better than the best silicon therapists that exist today. But I don't think that situation will persist any longer than such differences persisted in chess playing capabilities. Where for years I heard many people making the same mistake you're making, of saying that silicon could never demonstrate the flair and creativity of human chess players; that turned out to be false. It's simply human hubris to believe we possess capabilities that are impossible to duplicate in silicon.
A well-trained LLM that lacks any malevolent data...
The scale needed to produce an LLM that is fluent enough to be convincing precludes fine-grained filtering of input data. The usual methods of controlling an LLM essentially involve a broad-brush "don't say stuff like that" (RLHF) that inherently misses a lot of subtlties.
And even more, defining malevolent data is extremely difficult. Therapists often go along with things a patient say because otherwise they break rapport. But therapists have to balk once the patient dives into destructive delusions. But data of a therapy can't be easily labeled with "here's where you have to stop", just to name one problem.
It is an excellent idea - the first useful LLM most in finance have / will interact with is to throw the 1000's of daily reports into a vector database and query against that.
"Whats the consensus in todays research about AAPL?" Out comes a distilled report with clickable links back to the ai slop Goldmans et al sent out this morning.
> If you don't want people to feel pressure to socialize then you need to create a social norm of keeping to yourself.
This is the norm for any "outside locale" which does not assume socialization. For example, taking a walk on neighborhood sidewalks, visiting to a park, going to a local library.
> It would be nice for that to go away.
Pressure such as this originates from within. So to help that pressure go away, one must first embrace that socialization is a choice.
> So you advocate a costly post-harm remediation instead of a preemptive solution provided by governmental regulations
Absolutely. Such solutions tend to be much cheaper and more effective. It's how we deal with the vast majority of problems for a good reason.
If we sold oxycontin over the counter we would have much less of an overdose problem than we do. Would also take a lot of stress off our emergency medical care system which spends an inordinate amount of time just dealing with addicts looking for drugs.
It's a funny example to use to justify the current regulatory framework because oxycontin got approved by the very same.
>> So you advocate a costly post-harm remediation instead of a preemptive solution provided by governmental regulations
> Absolutely. Such solutions tend to be much cheaper and more effective.
This is not supported by any credible analysis I am aware of, as the cost of rectifying a problem post hoc has historically been far greater than preventing it in the first place.
> If we sold oxycontin over the counter we would have much less of an overdose problem than we do.
This assertion is demonstrably wrong and could easily be categorized as insulting to people struggling with OxyContin addiction.
People with opiate addiction have access to the cheapest and strongest narcotics ever available on the street. Unfortunately they are of uneven strength and cut with horrors that make their body shut down. I would much prefer they could just buy clean drugs from the pharma companies.
> People with opiate addiction have access to the cheapest and strongest narcotics ever available on the street. Unfortunately they are of uneven strength and cut with horrors that make their body shut down. I would much prefer they could just buy clean drugs from the pharma companies.
Fair point.
This is a complex societal problem having no simple solution. Unfortunately, what you proposed:
Should allow pharma to sell drugs to the market. Allow
people to sue if they are harmed. Do not care about a
government agency preventing access to new drugs.
Is not a viable solution either. Government agencies and regulations would be needed in order to ensure pharma's do not increase drug addictiveness or otherwise engage in nefarious activities. Agencies are also an avenue for accountability in distribution, a much cheaper and effective enforcement mechanism than civil lawsuits, and can provide options for people wanting to address their addictions.
Remember, every person who dies from an overdose cannot "sue if they get harmed" as they are dead.
They absolutely do. If for no other reason than each of their revenue goes to fewer entities.
> What you're seeing is the creation of wealth. This created wealth then flows out into the rest of the economy, via paying the workers and buying plant&equipment, etc.
This is the flawed reasoning behind "trickle-down economics"[0], which was called "horse and sparrow" in the 19th century. It didn't work in the 19th century when labelled as the latter nor in the 20th when reframed as the former.
Any company which has earnings beyond those of operating costs is a concentration of wealth by definition. Whether that wealth is distributed to shareholders, kept as retained earnings, or otherwise transferred to specific entities is irrelevant.
> Any company which has earnings beyond those of operating costs is a concentration of wealth by definition
Let's say I buy $20 worth of art supplies, and I paint a landscape and sign it with my moniker, "bright". Since "bright" paintings are very rare and go for a million bucks each, I now have created a million bucks of value. Who did I transfer the wealth from? Nobody. I took nuttin from nobody. Yet I have become wealthy.
Let's say I sell it to you for a million bucks. Did I take your wealth? Nope. I traded a million dollar painting for a million bucks. You are exactly as wealthy as you were before.
Now, if you decide to use my painting as compost (sob!), you are dissipating the million dollar value. That's not me concentrating wealth, it's you destroying your wealth.
If you stole the painting from me, then you concentrated wealth. But we're not talking about theft here.
Your art supplies and time didn't create wealth in the economic sense ("the annual produce and labour of the nation" as Smith defines it, <https://en.wikisource.org/wiki/The_Wealth_of_Nations/Book_II...>), but a financial asset, in the form of a (presumably) uniquely identifiable durable store of financial value.
This is contrasted with economic wealth in which some inputs (capital, labour, raw materials (themselves generally considered as capital) are transformed into some immediately useful consumable product (say, food or fuel), a durable good (clothing, furniture), a service ("immediately extinguished" in the terms of some economists, but still having potential lasting positive value), or capital which can itself be used in future economic activities (machining equipment used in the manufacture of widebody aircraft, or the aircraft themselves).
All else equal ("ceteris paribus"), the result of creating a novel but valued painting would be the same as that of inflating any other durable store of value: the relative prices of competing assets would fall proportionate to the price of the artwork. Ordinarily the drop is small and the bucket large such that this isn't noticed, but it is the net effect.
David Ricardo is amongst the first economists to deal with the economics of collectibles and durable assets of which I'm aware, e.g.,
Ricardo’s value theory applies only to those commodities that ‘can be increased in quantity by the exertion of human industry, and on the production of which competition operates without restraint’; it is thus not applicable to ‘rare statues and pictures, scarce books and coins, wines of a peculiar quality [...]’ (Works, I, 12)
From The Anthem Companion to David Ricardo, Edited by John E. King, p. 169.
>> Any company which has earnings beyond those of operating costs is a concentration of wealth by definition.
For context, here is the second sentence related to the above:
Whether that wealth is distributed to shareholders, kept as
retained earnings, or otherwise transferred to specific
entities is irrelevant.
I believe this relevant to the below.
> Let's say I buy $20 worth of art supplies, and I paint a landscape and sign it with my moniker, "bright". Since "bright" paintings are very rare and go for a million bucks each, I now have created a million bucks of value. Who did I transfer the wealth from? Nobody. I took nuttin from nobody. Yet I have become wealthy.
This scenario is not relevant to "big business", but instead describes a sole proprietorship with an assumption of a known fungible value. It also does not account for consumable goods and/or transient services. In any event, by your own definition below:
> Let's say I sell it to you for a million bucks. Did I take your wealth? Nope. I traded a million dollar painting for a million bucks. You are exactly as wealthy as you were before.
Either you have not "become wealthy", as you "traded a million dollar painting for a million bucks" or the effectual value of money exceeds the purchase value of the "million dollar painting". Both cannot be true.
Back to my statement of corporate earnings beyond operational costs being a concentration of wealth. I believe we can agree on the following:
- Any for-profit company has as its purpose the goal of accounts receivable (AR) exceeding accounts payable (AP) over time.
- There are a limited number of recipients regarding profit distribution for any given company.
- Those having no direct or indirect ownership of a given company do not receive profit distributions.
If we agree on the above, then the larger the profits, the larger the distributions. Since the set of people qualifying for profit distributions are less than the set of all people having no investment relation to the company, it follows that said profits are enjoyed by fewer entities than those strictly involved in the AR side of the ledger.
Hence a concentration of wealth into the organization.
> Either you have not "become wealthy", as you "traded a million dollar painting for a million bucks" or the effectual value of money exceeds the purchase value of the "million dollar painting". Both cannot be true.
I became wealthy by creating wealth, not concentrating it. Concentrating it requires it be taken from somewhere else. There is no taking going on, there is creation and exchange.
>> Either you have not "become wealthy", as you "traded a million dollar painting for a million bucks" or the effectual value of money exceeds the purchase value of the "million dollar painting". Both cannot be true.
> I became wealthy by creating wealth, not concentrating it. Concentrating it requires it be taken from somewhere else. There is no taking going on, there is creation and exchange.
The scenario you have described is logically consistent while being representative of a tiny subset of commerce. Revisiting the original use-case:
Let's say I buy $20 worth of art supplies, and I paint a
landscape and sign it with my moniker, "bright". Since
"bright" paintings are very rare and go for a million bucks
each, I now have created a million bucks of value.
Assuming all of the above, this business model does not account for at least the following:
A - Businesses having more than one employee.
B - Asset deprecation, such as when purchasing a new automobile.
C - Consumable goods, such as food, petrol, etc.
D - Services such as commercial/residential rent and physical security.
E - Taxes.
F - Stock dividends and/or performance bonuses.
A and E involve direct wealth transfer from the business to relevant parties.
B is a second order effect only realized when the buyer attempts to sell the asset to a third party.
C and D are direct wealth transfers as the seller retains the remuneration for as long as they desire (excluding applicable cases identified above) and the buyer eventually does not have a physical equivalent. Note that this often remains a valuable exchange for both parties.
When a company outsources their core competencies in the pursuit of reducing costs, it inevitably becomes wholly dependent upon its vendors and loses its previous capability to independently deliver. Most companies which choose this path either fail or are purchased by competitors.
The same can be said for individuals whom outsource their understanding of both what must be done and how to do it to a statistical text generator.
> As a long-time compulsive Internet user, I am aware of the emotional and psychological risks of this new technology.
> Are there any technologies or apps that are worse than others, particularly for people with obsessive/compulsive tendencies?
Social media, gambling, and "freemium game" sites/apps all qualify as worse than LLM-based offerings in the opinions of many. Not to mention the addictiveness of their use on smartphones.
However, the above are relative quantifications and in no way exonerate LLM offerings.
In other words, it doesn't matter how much poop is atop an otherwise desirable sandwich. It is still a poop sandwich.
> There used to be a contest to fit a good web page into 5kB ...
In accordance with this philosophy, I worked on a project a couple years ago where the HTML home page uncompressed size was intentionally limited to 4k. The idea being a slow network connection (such as a mobile device using a limited cellular connection) would be able to render quickly and remaining content could load asynchronously.
How about having a page consisting of pure html, less than 0.5kb, with the server only providing content alone (simple, one mp4), but serving you none of the html?
So it even works when there is no page on the web at all :0
In that case your browser only needs to load 0kb of html from the web in order to successfully reach the page.
That can be pretty fast. Either way. 0.5kb is real small but 0kb is a bit smaller ;)
Fairly elementary, all you're wanting to do is access some media which is found on the web itself at a known address.
But there's a catch, you would have to supply the html "landing page" file yourself from your own PC if none of the html's going to be coming in from the web to your browser.
Copy and paste the code into a text editor, save file as fuzzplayer.txt into your favorite folder, then save it again as fuzzplayer.html, right there in the same folder. To edit in the future, double-click on the TXT version to edit, double save again as both file types when done. Double-click the identical html version (identical to txt except for file extension) to launch your browser and go to "my" page where you can play the media. Not much differently than the Mozilla example does.
"Header" (not formally), three "paragraphs" (without paragraph marks), EOF.
Backward indenting for me, old (without) school I guess.
Pretty straightforward
Almost everything indented except things that you might want to change more easily in the future. Then you can just go down the left margin and kill 'em all or take pot shots.
Page formatting is associated with a block of displayed text in preference to other objects. This is the indented html formatting which is quite likely to need some future editing, very often in concert with the text itself, so it's good to have it right there. Almost seems like some html belongs without indentation also but that kind of defeats the purpose of the "clean" look. You get used to it.
Editing 5kb or 10kb of this kind of stuff manually, covering all the bases in a couple minutes, and having it work the first time can make you smile more than a lot of things :)
They are very useful algorithms which solve for document generation. That's it.
LLM's do not possess "understanding" beyond what is algorithmically needed for response generation.
LLM's do not possess shared experiences people have in order to potentially relate to others in therapy sessions as LLM's are not people.
LLM's do not possess professional experience needed for successful therapy, such as knowing when to not say something as LLM's are not people.
In short, LLM's are not people.
reply