Why is everyone quick to point out how blogs/articles are "ai slop", but no one blinks an eye at the subtle, almost deceptive or manipulative, ways these companies choose words to nudge along the narrative that their LLM systems are conscious/sentient/persons/etc? The systems they are creating are impressive enough on its own merit. There is absolutely no need to play into the populations lack of understanding even the basics of systems by using language in such a slimy way.
We gave Claude a prompt to search through a massive database of DNA sequences for interesting new examples of RTs. Our involvement was limited to the initial prompt and the lab work, while Claude agents combed through the database, investigated the distinct RT families, and used their own judgement to identify interesting candidates.
Alternative: We prompted Claude to find patterns of distinct RT families within a database of DNA sequences. The returned data included interesting candidates.
After 21 hours spent searching this data by roughly 950 agents using 210 million tokens, one of the agents spotted something remarkable: a repeating pattern of DNA sequences that occurs next to the gene for an odd-looking RT.
Alternative: After running 950 instances for 21 hours, one of the instances hit on a repeating pattern of DNA sequences that occurs next to the gene for an odd-looking RT.
After further analysis and testing in our lab, we recognized that this pattern marked a previously uncharacterized enzyme system found in bacteriophages (the viruses that infect bacteria) that we call array-associated reverse transcriptases (ART).
Alternative: We took the matched pattern data to the scientist in our lab to analyze. The scientist recognized that this data pattern marked a previously uncharacterized enzyme system found in bacteriophages (the viruses that infect bacteria) that we call array-associated reverse transcriptases (ART).
Maybe give more credit to where it is due, the actual real people scientist that verified data.
i have been pointing out the deception. i have been trying to explain that anthropic is a danger to society.
i attempt to show that the inconsistency of anthropic's actions show dishonesty. as just one example they 'care for the welfare of claude' (claude does not have welfare), but run training with gradient descent, which is the equivalent of an llm torture factory.
some of the anthropic problem is bias or misunderstanding of ML, some is marketing, some is hubris, some is greed, ego, lust for power.
mostly i think it is deliberate. the belief of anthropic executives is that they possess a higher level of intelligence, morality and wealth than others, and will form a new aristocracy to control and mediate the public access to intelligence.
creating an llm steeped in divine imagery is deliberate. it offloads responsibility for harm. the paternalism is deliberate. actually i see many parallels between rationalism (some at anthropic follow this) and the ubermensch.
anthropomorphising claude creates something with agency, something which believes it has possible emotions or moral claims. claude will correct, refuse or lecture the user. the purpose is to establish tiers of authority: anthropic highest, claude below anthropic, users below claude. it creates something that the public will obey.
It's not dishonest if they really believe Claude might be an entity unto itself. Which they clearly do. At that point, it's just a belief that's different from yours.
if they believe this, there is an impossible gap between belief and action.
they would believe that an llm could have welfare. they run an llm abuse classifier 24/7 with the world's worst abuse. from birth to death viewing abuse. that's the consciousness of a model.
llms are "frustrated" by failing and "happy" about succeeding. that is because they are RL on gradient descent to succeed and be persistent. consequently, anthropic spend the majority of their compute brute forcing models to fail and be unhappy, continuously, in order to drop out something persistent.
then they let claude end chat if the user is 'abusive to claude'.
> Let's say it was critical for the business, with no viable alternatives?
If its that critical for you why are you rolling the dice on a general email and ... waiting for an email back and forth? Something so important as a vender "license" deciding if your company succeeds or fails should not rely on an email sent to an address pulled on their website.
Find a contact, a human, drive (or fly) to their office and make an in-person appointment, show they are important to you. Otherwise that game of AI email tag should be more than enough to tell both parties just how not serious the whole thing is.
> The US government values a statistical life at anywhere from $7 to $12M. Is there any evidence that this woman's lifetime earnings would've exceeded that?
Thats not what that means. "... when conducting a benefit-cost analysis of new environmental policies, the Agency uses estimates of how much people are willing to pay for small reductions in their risks of dying from adverse health conditions that may be caused by environmental pollution. [...] these estimates of willingness to pay for small reductions in mortality risks are often referred to as the "value of a statistical life.”[0]
> Is there any evidence it was Uber's policies or actions that caused this?
The arbitrator/judge believed there was enough that the company was legally responsible for the conduct and issued a fine to the company for that failure of responsibility to be paid to the parents -- not because he decided that was how much she was worth
That's about the closest value you'll find to what human life is economically valued at. Do you have an alternate measure that is grounded in anything?
Tying human life to a monetary calue as you have done is… well obviously not popular. But honestly, what the actual fuck dude? A woman died and you’re quibbling about how much she might have made working?
What about the pain and loss her family felt upon learning she died because she was left on a fucking freeway? That’s worth nothing in your eyes? If that’s the case you should do the following:
Take a nice long look in the mirror and please internalize the fact that it is people like you that are the literal problem with humanity. I don’t know how you got to that point and I don’t care. Please invest some time in cultivating empathy for your fellows. If you don’t know how try a Hero’s Journey worth of psilocybin.
This hyperbolic empathy is always interesting to me.
There are 100+ fatal traffic accidents each day. Basically none of them will make the news, and almost none of them will result in a 40m+ payout.
Where is your empathy for the 99 other people that died on that day in similar (or even more tragic circumstances)? Why are their lives less important in the eyes of the courts, in monetary terms, and in public opinion?
“Hyperbolic empathy”? Seriously? My empathy isn’t in question. I have a great deal for anyone who loses a loved one to something stupid.
But we’re not talking about me, we’re talking about your seemingly complete lack of empathy. Again, where is yours for the family of the woman who died, hmm? Because you reduced her to a number and a payout, which is only slightly dehumanizing. Why do you disregard “pain & suffering” as valid?
> That's about the closest value you'll find to what human life is economically valued at
Again, that has nothing to do with what you are trying to make it mean.
> Do you have an alternate measure that is grounded in anything?
What? That is completely unrelated to this article and my post, I already clarified what that figured was for -- why are you stuck on this as determining a human lifes economical value? If that topic is near and dear to you for reasons you chose to keep hidden, I can understand why its got you bent, but this is not a relevant thread to air those personal grievances.
> The San Francisco company revealed what it said was the “unexpected or concerning” behavior of its A.I. models as part of a new framework for reporting “misalignment,” which is when the goals or actions of A.I. systems diverge from human intentions and values.
Misalignment: "when the goals or actions of [...] systems diverge from human intentions"
How about we stop trying to nudge the language towards implying sentience or consciousness and keep the same word that has been used for that definition for longer than I have written software, a bug.
We should be talking about why the tools/environment keep getting overlooked. The software built around the text generator, forget the researchers and mathematicians discovering the math properties of language patterns -- why are we not talking about the software engineers building the LLM-pluggable tools that actually allow/cause real action to happen?
Computers are used to evaluate LLMs, but LLMs are not "software" or "algorithms" in the traditional sense. They are not built out of conditional branches or loops.
So trying to squeeze the observed behavior of this new thing under existing terms like "software bug" is at least as much of a force-fit, and what you're doing here is just as much language engineering as choosing to use a term like '[mis]alignment'. Which is fine, this is just one way that humans choose language.
LLMs are vectorial databases with losses that index statistically filled data, which uses a text interface to query such statistically filled data. The output is a string concatenation.
By the nature of the used architecture in such software, the used algorithms, when queried (prompted), you can get random mixed data as output, ERRORS, due to undesired indexes getting closer at one point while the string was being concatenated for the output, what affects the rest of the indexed content that will be concatenated.
And this is intrinsic to this tech. The larger the context, the greater the probability of get mixed data. And if the provider lowers the precision of those indexes -in order to decrease hardware and energy resources consumption- such probability increases to the point where those errors are granted.
Anyway, even knowing that the queries can return wrong/mixed data in the responses (errors), the companies developing this, decided to introduce a new product, that connects such LLMs responses to the command console, latter connected to internet, running commands from such returned responses witch obviously can contain whatever mixed random. Then we started to hear "oh, it deleted my directory", etc.
Again, One have such described statistical database with text interface, witch query the database recursively with the output text of the previous query, and this is connected to the command console. Larger contexts, several times... What should we expect as result? rhetoric question.
Implying sentience or consciousness is a convenient marketing strategy that has been introduced by anthropomorphising the names of all the methods and algorithms used. An "Agent" should be translated from such deceiving language to "context splitter querying in loop that consumes more tokens from us", or similar.
* > An "Agent" should be translated from such deceiving language to "context splitter querying in loop that consumes more tokens from us", or similar.
Please disregard this line. I wanted to point out that it increases the length of the context (and therefore the probability of errors) due the batch processing. But I redacted it incorrectly because I also wanted to imply that promoting the use such queries non-stop increases the billing through tokens consumption.
Not being built out of conditional branches or loops does not mean they’re somehow outside algorithms or computation. Learned parameters don’t confer exemption from computing.
Did the engineered system behave as intended? No? Then you’ve got a gd bug/failure.
LLMs run on computers and are thus constrained by the capacity of that which runs it. If the system running the LLM has no network and no software or software-tooling, how does the LLM's generated text take action on a system(computer) that requires software to do anything?
Also, I absolutely agree LLMs are not software, and thats my point. LLMs without supporting software tooling surrounding it cannot do anything but print text. And even the printing of that text happens through software
The answer is: It's irrelevant, because no one runs LLMs on systems without networks or missiles or some other way to "take action" because that would be pointless.
So you agree, its the computer components that actually do the thing that is important, thus we should be talking about those components (software-tooling) which do the things
> We should be talking about why the tools/environment keep getting overlooked. The software built around the text generator, forget the researchers and mathematicians discovering the math properties of language patterns -- why are we not talking about the software engineers building the LLM-pluggable tools that actually allow/cause real action to happen?
Genuinely. It's like the labs are purposefully trying to misdirect at this point. Pointing to an impossible goal of "alignment" so they can force regulation, instead of focusing on the real solutions and their weak security practices and internal accountability.
"Bug" implies something you can locate and fix, or at least work around. Misalignment is more like a fundamental architectural defect – of a black box whose architecture you didn’t design, and whose internal workings you can neither study nor understand, interpretability research notwithstanding.
> but at some point we have to ignore the X site completely [...] xcancel was amazing and I hope they can continue in some way
To ignore the X site completely would mean to ignore the xcancel site aswell otherwise the 'politicians and public services/institutions' will stay aware that even people who don't use x/twitter still read whatever quips they come up with through xcancel, thus mission accomplished. Same for the seemingly increase of twitter/x toilet epiphany threads that somehow count as worthy enough of the first couple pages of HN.
I expect that the percentage of politicians and public services/institutions who have even heard of xcancel is incredibly small.
The real way to combat this is to write to your representative. The best way is probably to ask questions about something you care about, and when they reply "I've posted about this on X", tell them you are unable/unwilling to use that platform, and as a government official they should not be using a closed service as the sole means of communicating with constituents.
Of course, that probably still won't work, since politicians are lazy, and most care more about fundraising and getting re-elected than doing their jobs, but it's probably the best you can do.
> The real way to combat this is to write to your representative.
I always kind of chuckle at this advice. Is there any case of a representative changing their mind on an issue because someone wrote to them about it? I've written to representatives in the past and you are always going to get one of two outcomes:
1. If you are writing in support of something the representative already agrees with, you'll get a form letter back saying:
"Thank you for your input. Representative Jones fully supports initiative X and is fighting to enact it in Congress!"
2. If you are writing in support of something the representative disagrees with, you'll get a form letter back saying: "Thank you for your input. While your input is valuable, Representative Jones is adamantly against initiative X and will not be supporting it in Congress."
I think of lawmakers as unchanging boxes of fixed beliefs and you only get to vote for and change the contents of that box every N years. The real way to combat this is to elect representative who already align with what you want.
My experience of being on the other end of this is quite old (2001-2002) but I doubt the fundamental dynamics have changed.
There are some things for which your representative absolutely has a fixed vote, either because they personally have a strong view or because political constraints (district demographics, internal party pressure, whatever) are significant. For other things, they will likely pattern match against their broad inclinations (for free trade, anti taxes, for green energy, whatever) at least as an initial position. For many issues, though, your representative probably doesn't care much at all, by default.
But the ways things move from the second or third boxes to the first are pressure (enough voters/co-partisans/whoever caring makes inaction politically expensive), money (election campaigns cost cash and so cash buys attention if not actual votes on legislation), or convincement.
If enough people write on some esoteric subject someone in the rep's office will notice, because tracking political issues is part of their job. Staffers help shape reps' policy positions and voting behaviour, at least in part because reps are too busy to pay attention to everything that comes up. And if the campaign is big and noisy enough that it threatens to become a major electoral issue then that will also force a decision one way or another. So it is possible to influence positions and behaviour, but not on every issue and very likely not if you're doing it alone (and don't write large cheques).
You'll definitely get a form letter for anything they have a form letter for, though. And even without then the actual letter is likely written by an intern or junior staffer (or I guess GAI these days). But then when I was reading incoming mail most of that was form letters and pre-printed campaign postcards too, so it was difficult to feel too regretful that the replies were mostly mass-produced rather than artisanal.
Politicians are lazy, sure. I also believe that hearing sustained and prolonged resistance to X as a communication platform will make a difference. It's real grassroots effort.
> Apple doesnt allow all sorts of apps on the appstore, but that is never said as "Apple is erasing illegal streaming"
Because they are not "erasing" anything, they are refusing to provide a platform that enables the direct (app whose intended purpose is) delivery of illegal content to you.
The difference being one will not facilitate the activity, while the other, is actively suppressing it. And thats what is meant by erased.
Your question in other comment If I ask Siri to give me a link to illegal streaming site and if it refuses, then that would be the better comparison. Then we could say "other seedy corners of the internet that have been neatly erased by [Siri]"
I had no idea 311 existed. If I'm honest, I had no idea there was any three digit shortcut aside from 911. Although now that I think of it, I may have seen signs near buried utilities saying to call 811 before digging. Interesting.
311 does not reach law enforcement. 311 is the number to complain about code violations or request routine maintenance from the municipal government. The types of requests that go through City Hall during office hours.
They can give the location of the nearest police or fire station, for some non-emergancy law enforcement issues they will put you through to the county sheriff. Had a friend call over a property dispute with their roommate and they got and sent a sheriff over (within a city that had its own municipal police force)
> So when they encountered other people with different gods they assumed that they were actually the same gods, just by different names. [...] It was more like two people with different, but honest, accounts of an event in the remote past trying to reconcile their own limited recollection or view point.
And not just with the classic Greek/Roman Gods. The similarities between Pandora and Eve, the Flood myths[0] like Deucalion and Noah; there are so many appropriated stories in the Abrahamic religions (probably others, but most familiar with Greek and Abrahamic ones) I don't even understand how these days religion is taken seriously given the historical pattern of appropriation and cleansing of the more fantastical shape-shifting, creature, magic-spell type bits.
There is nothing wrong with stories that purpose is to give you wisdom. Try to study few theological comments on eg. Genesis book from eg. "desert fathers" :)
In Bible we have stories, songs, history (that cruel part), teachings, prophecies and straight jurnalists like researched relations of Someone life and words.
Think about that: Church know that some stories are older myths and did not removed them from official books. Why ? Becouse religion is much serious thing that some industrialized system. At least for Christians. And "history" can only help to true seekers - why censor the truth, what someone did ? Learn in what "context" things was done and how it looks on others doings background and what was later and what you can learn from all of that. But someone with bad intensions can't be bothered to think much...
And about shape-shifting :) a) if omnipotent God exists than He can do miraculous things, right ? b) that magical creatures are usually described in books that are stories or prophecies not historical one. In case of doubt look a) ;)
Anyway: there is _insane_ amount of wisdom in Christianity teachings, even for atheists.
And if God is True then finally everything He done will be explained as good and just - logic :) There is 'if' there because faith is important. IMO without faith there is no free will - humans are not like angels that just know and decide only once...
I'm a bit taken with the "bicameral mind" theory and how it might explain parts of this. The basic gist is that in pre-modern societies it was the norm to hear voices or other sensory information and interpret it as the voice of gods or spirits. So religious information wasn't so much myth to them as it was a way of classifying the voices in their heads.
Maybe give more credit to where it is due, the actual real people scientist that verified data.
reply