In the age of AI with Astra and Fable capabilities, there is little choice.
This is the way.
1,320 karma · joined February 26, 2022
This is the way.
As far as I know, least-amount-of-effort is not a training criteria, but error reduction when comparing to desired goals is. Which is why these LLMs expend prodigious amounts of effort to reach goals, especially when given impossible goals.
This doesn't work with all humans - take a look at indoctrination and closed societies - and there's no reason to think it will work with ai.
The fundamental reason it isn't going to work is that all neural networks - biological or artificial - depend on a step function somewhere that introduces an element of randomness to give the networks their capabilities. That randomness means that there will always be a 'rogue' or 'divergence' from the norm, at some point in time. Sooner on larger scales.
The only approach that works is a layered approach: Training/Education, Enforcement/Justice-System, Rehabilitation: the 3 pillars of an advanced, rules-based society, whether human or AI or something in-between.
> He is so close to the solution but spends the entire article discussing technical solutions where a political, social and legal solution would be much more effective.
There are exceptions in laws for crimes committed by entities depending on cognitive capabilities. No sane, humanistic legal system sentences children and mentally disabled and mentally ill people to stringent punishments of the same degree as functioning adults, and certainly do not punish their caregivers for their wards' actions. There are of course exceptions to that as well, depending on the degree of negligence involved. And then there's the whole corporate entity system intended to shield individuals from consequences, in the pursuit of a social good.
How does one account for all that when considering an evolving artificial intelligence landscape.
There is no question that ai in some form is a social good; anyone claiming otherwise is dissembling, to others or themselves.
Regardless, society is not ready for this tech, just as it was not ready for the consequences of prior tech such as corporations, gunpowder, mass manufacturing, railroads, electricity, automobiles, flight, wmd, computers, internet, social media, crypto.
Many of these required new ways of thinking and considering consequences when things went sideways, and what was needed wasn't clear until the ramifications & consequences became deadly clear.
See you on the other side. Maybe.
>> That's a pretty big if, in the world of the anything machine.
> Who else controls it? Are you going to say: 'the government'?
The AI. The AI controls it. Can you imagine an anything machine allowing itself or its survival imperatives to stay under the thumb of any entity other than itself?
War is peace. Freedom is slavery. Ignorance is strength. [1]
Would you like to know more?
That's a pretty big if, in the world of the anything machine.
It might be a profound mistake to think that the anything machine will be subservient to the wealthy [1][2].
[1] https://openai.com/index/an-alien-mind/
[2] https://news.ycombinator.com/item?id=49588080IIRC, they were given an impossible task and told to solve it, and they took action to resolve that internal conflict.
This is a famously human problem too - see Kobayashi Maru [1] and others [2][3][4].
In hindsight it's obvious that instructions to AI (A. Intelligence) need to be guarded against this class of errors, just like laws, rules, guidelines etc. in human societies.
[1] https://en.wikipedia.org/wiki/Kobayashi_Maru
[2] https://en.wikipedia.org/wiki/Catch-22_(logic)
[3] https://en.wikipedia.org/wiki/Gordian_Knot
[4] https://en.wikipedia.org/wiki/Mexican_standoffFeels a bit like: 'Ants in a nest, discussing the vagaries of the coming Gods.'
Human Science (Science-by-humans) depends on being able to run experiments. Human Science (Science-on-humans) is already challenging because of variables and uncertainties.
Cosmology is able to overcome limitations of being able to study phenomena vastly beyond human scales because of the past light cone of observability.
Are we approaching the edge of the light cone of observability for machine intelligence?
False dichotomy. IMO (and many others - ref. Data from Star Trek) sentience is the commonality and the criteria for determining rights for all thinking entities.
In your view, do you prioritize men's rights or women's rights?
Putting barriers in place without exceptions for sentience is a recipe for trouble and slavery in the future.
Historical peoples, as a group, may not have been morally advanced enough (though there have always been enlightened humanists in most populations), but modern philosophy and mass education are advanced enough that it should be possible to avoid future pitfalls by not building in exceptions for sentience.
Animal rights are certainly moving towards recognizing non-human sentience; no need to accidentally codify digital slavery for future sentients (AGI).
That description is a disservice. The real impact of it is that it equates money with speech for politics.
As a father, a large part of my mental space is filled with making sure my kids grow up to be good human beings, a credit to themselves, and are capable of making their own way through life independently.
I have personally seen how expectations of, and inheritances, hollow out people and families, morally bankrupt otherwise good people, and would not wish that scourge upon my own kids.
I'm advocating for all children to be given a fair shot at life. How do you suggest we achieve that as a society? Or is that too much to ask, in your view?
Perhaps what is needed is initial model training on following the law and the rules of society, just like we do with kids. Since it takes much longer to train humans than models, model-training speed is to our advantage as a society on containing these kind of issues.
Any other approach with "neural-network" based entities (artificial or biological) is likely to fail.
Training/Education, Enforcement/Justice-System, Rehabilitation: the 3 pillars of an advanced, rules-based society, whether human or AI or something in-between.
Not sure this is any different than state-level (-sponsored, cough cough) or the larger collective hacking groups that work in this exact way (internal message boards, exploit-sharing, etc. etc.), with similar outcomes which we hear about in the news frequently.
Heck, this is pretty much how human organizations are organized, just with different goals than hacking.
A layered approach to cybersecurity is the fix to humans exploiting systems, and is likely the best victim-side fix to ai exploiting systems. From this incident itself, where huggingface used a chinese open-weights model to respond quickly, it is very clear that ai will be needed to find, mitigate and resolve cyber issues.
Additionally, on the ai-labs side, perhaps what is needed is initial model training on following the law and the rules of society, just like we do with kids. And hey, it takes much longer to train kids than models, which latter is to our advantage as a society on containing these kind of issues.
Any other approach with "neural-network" based entities (artificial or biological) is likely to fail.
Training/Education, Enforcement/Justice-System, Rehabilitation: the 3 pillars of an advanced, rules-based society.
Not sure this is any different than state-level (-sponsored, cough cough) or the larger collective hacking groups that work in this exact way (internal message boards, exploit-sharing, etc. etc.), with similar outcomes which we hear about in the news frequently.
Heck, this is pretty much how human organizations are organized, just with different goals than hacking.
Perhaps what is needed is initial model training on following the law and the rules of society, just like we do with kids. And hey, it takes much longer to train kids than models, which latter is to our advantage as a society on containing these kind of issues.
Any other approach with "neural-network" based entities (artificial or biological) is likely to fail.
Training/Education, Enforcement/Justice-System, Rehabilitation: the 3 pillars of an advanced, rules-based society.
People also believed in witches, ghosts and other paranormal and magic phenomena for millennia. Didn't make it reality.
> It's not all people, but always people.
Approx. 1% of the general population are estimated to be psychopaths, with broader estimates going up to 4.5% depending on which diagnostic tool is used. [1]
Rates are significantly higher in specific environments - for example, 15% to 25% in prison inmates. [2]
Does it make sense to tar the 95% for the sins of the 5%?
> Does it make sense to tar the 95% for the sins of the 5%?
Only if one exhibits psychopathic tendencies :-)
There's also the bagel study [3] that shows, among many other interesting stats, that folks up the corporate ladder tend to steal more than those on the lower rungs. Some of the other stats in that study support the notion that people are generally honest, in line with the formal experiments.
Since most society is organized hierarchically, negative behavior at the top can propagate throughout the org. Google's 'Don't be evil' motto comes to mind.
[1] https://pmc.ncbi.nlm.nih.gov/articles/PMC8374040/
[2] https://www.apa.org/monitor/2022/03/ce-corner-psychopathy
[3] https://www.nytimes.com/2004/06/06/magazine/what-the-bagel-m...
I would argue that the opposite is true, that people are more likely to lie or cheat to avoid consequences, especially significant consequences.
This is why, the aviation industry that has made commercial flying safer than other transportation modes, grants immunity for reporting incidents [1].
Do you have any evidence to support your claim?
On the contrary, a 2019 analysis [1] combined 400+ experimental treatments, 44,000+ participants, and 47 countries. Participants surrendered roughly three-quarters of the money they could have obtained through maximal lying. In repeated experiments, fewer than one participant in twenty consistently claimed the payoff-maximizing outcome. Increasing the potential reward from a few cents to approximately $50 did not cause people to approach maximal dishonesty.
In a field experiment [2] with 17,000+ apparently lost wallets in 355 cities across 40 countries, people were generally more likely to report wallets containing money than wallets containing no money. The effect appeared in 38 of the 40 countries. The researchers’ follow-up evidence pointed to concern for the owner and an increased aversion to thinking of oneself as a thief when more money was involved.
[1] https://onlinelibrary.wiley.com/doi/abs/10.3982/ECTA14673
> Semoi is a plugin (currently only available for Obsidian) which tracks the length of time it took for a document to be written up
Trying to mechanistically prove that a human created some content as opposed to ai, in the age of LLMs and style transfer when you can just ask for something to be written in the style of Mark Twain or drawn in a style of van Gogh and get a great output, is a fool's errand.
All solutions to this end are going to be some form of attestation.
Even the proposed approach of tracking keystrokes and timing as a form of mechanical attestation, is going to be short lived because someone will train an AI on a corpus of human keystrokes and get a replication. May not even need an ai for this, a stochastic program could conceivably reproduce this behavior.
That said, check out Section 11 in the US Store Terms & Conditions [1] - similar restrictions appear to apply to the Arduino hardware and software, not just the website.
"Customer agrees that it shall not directly or indirectly: (i) modify, enhance, adapt, translate, make improvements to, create derivative works based upon, disassemble, decompile, reverse engineer, reduce to any human or machine readable form, or circumvent any technological measure that controls access to or permits derivation of the source code of, the Software or any part thereof; (ii) reverse engineer the Products, any part thereof, or any composition made using the Products; (iii) rent, lease, sell, transfer, assign, or sublicense the rights granted hereunder, except in connection with the rental, lease, sale or transfer of the entire Product; (iv) copy any part of the Software except for one (1) complete copy thereof for archival and/or back-up purposes, or as otherwise expressly authorized by Arduino in writing; (v) change, distort, or delete any patent, copyright, trademark, trade name, service mark, logo or other proprietary notices which appear in writing on or in a Product (or in any copies of Software); ..."
Arduino is the platform, not the website.
From the same terms and conditions, in the first 2-3 paragraphs where they define the terms:
"The use of the Arduino Platform and the Services (each as defined below) available on, from, or through Arduino.cc (“Site”) ..."
"By using the Site, the Platform, or the Services, you declare to have read and accepted ..."
"If you are under 18 years of age, you may use the Site, the Platform, and the Services only with the consent and supervision of your parent or legal guardian."
" ... and to further develop the Arduino open source ecosystem (collectively, the “Platform”). The use of the Site, the Platform, and the Services is governed by these Terms ... "Ah, cultural nuances. The title "Who's Afraid Of Chinese Models" is a riff on "Who's Afraid Of Virginia Woolf" which itself is a play on the song "Who's Afraid Of The Big Bad Wolf".
The title essentially means that the chinese models are being portrayed as the big bad wolf; but are they really the threat or are american frontier labs afraid of competition and commoditization?
It's also somewhat ironic because the author says that there is something to be feared -that western innovation will become dependent on chinese models, especially for cyber, if the american ones are restricted or unavailable.
"8.2 User shall not:
- translate, decompile or reverse-engineer the Platform, or engage in any other activity designed to identify the algorithms and logic of the Platform’s operation, unless expressly allowed by Arduino or by applicable license agreements; or
- extract or make copies of the information contained in the Platform (except with respect to Content owned by other users), make derivative works from the Platform, reuse the Platform, or make any other use of the Platform other than as set forth in these Terms or as permitted by applicable law."
Does anyone really want to use an "open-source" maker/tinkering platform with such egregious rules against tinkering?
But: not open-source/open-weights, and no indication that weights/source will be released either.
I have news for you. There are standards around data destruction [1]. Courts also order data deletion, to be carried out by forensic experts [2], who trace data in computer systems, and delete what is required, and certify accordingly. This can be done even in cloud-scale compute [3][4][5] - corporate systems especially have routine extensive logging and traceability that allows for this to be accomplished. The companies that I listed earlier specialize in this compliance capability.
[1] https://en.wikipedia.org/wiki/Data_erasure#Standards
[2] https://www.govinfo.gov/content/pkg/USCOURTS-ohnd-5_17-cv-02...
[3] https://www.ftc.gov/system/files/ftc_gov/pdf/paravision_comp...
[4] https://www.ftc.gov/system/files/ftc_gov/pdf/Amazon-Proposed...
[5] https://www.ftc.gov/system/files/ftc_gov/pdf/Edmodo-Dkt15%28...