[Xmca-l] Re: The ethics of artificial intelligence, past present and future

[Bill Kerr]

You have to wonder what computers would be like now if Rosenblatt had been
able to pursue his work unfettered by Minsky and others from MIT back then.

Academic freedom must be protected. On that I hope we can agree!


Perceptrons was written by MInsky and Papert in 1969. Many have argued and
I agree with them that their other work kicked off an extremely rich field
of educational computing (call it the MIT group if you want), which
persists in numerous branches today: Scratch3.0, AppInventor, Makey Makey
all came out of MIT not to mention the associated theoretical work.

I google Perceptrons and it confirmed what I thought from before: that it
made legitimate criticisms of  that path of research. It was a  legitimate
dispute between different approaches at that time. I can't evaluate myself
because my maths isn't good enough.

It is true that work on Perceptrons dried up for quite a while after that
probably because no one could refute the critique by Minsky and Papert.
Irrespective, in your words Annalisa we should support their academic
freedom to argue their case.

It is also true that  parallel processing perceptrons / neural networks
have achieved remarkable things in recent years. It seems  that MInsky /
Papert made a legitimate criticism which ended up sidelining what turned
out to be another rich research field.

Decades later Minsky, now dead, is accused of having sex with a 17yo at
Epstein's compound when he was in his 70s.

Therefore, what "Minsky and others" did before that is now suspect by
association. Is that what you are arguing Annalisa?




On Sun, Dec 22, 2019 at 2:37 PM Annalisa Aguilar <annalisa@unm.edu> wrote:

> Hi Ed,
>
> Regarding Dreyfus, I don't recall him asserting the matter of mind or not,
> though it's been almost 10 years since I read the book.
>
> I am compelled to say that minds are material in the same way that stories
> are material.
>
> Consider a few analogies.
>
> The book is material, the words are printed ink on the paper of the pages,
> but without the book present the story will not manifest in the mind of a
> reader (as long as the book is written in the same language as the reader).
> Is the story not material if it is located in this book and not in that
> one? Also the story can exist outside the book, in the memories of a
> person, but the person is also material.
>
> The light in an electrical light bulb is there when the electricity passes
> through the filament, and not when the electricity is not there. We know
> thanks to Einstein that light is energetic material that travels really
> fast. The filament is gross material, the electricity is subtle as is the
> light, but the three are material.
>
> I assert that a mind too is subtle energy passing through a brain, which
> is a conglomerate of neuronal connections of grey matter.
>
> I see the physical and transactional world as material of infinitely
> different graded properties, subtle to gross, in different combinations of
> active qualities. In the same way the story resides in the book and the
> light resides, or emanates, from the light bulb the subtle permeates the
> gross.
>
> A more perfect illustration is the red hot iron ball. Iron and fire are in
> the same location, one is gross the other subtle. But both are material.
> What can happen however is if we do not know the properties of iron (heavy
> and round) or fire (red and hot) we can superimpose one element upon the
> other (i.e., assert that fire is heavy and round, while iron is hot and
> red) and this is easy to do because they are present in the same location
> perceptually; we cannot remove the iron from the fire or vice versa.
> (though it is possible if you are a blacksmith you can purge the iron in
> water, extinguishing the flame, I suppose, but you get my point, I hope.)
>
> With this in mind, is it possible to also assert that ethics is also a
> material entity? Whereby ethical conduct is that which possesses the most
> truth for the most harmony for the largest part of society while also
> holding the same for the individual.
>
> Can ethical conduct have universal laws like physics? If so, it might be
> an attainable goal to create the ethical algorithm. Yet, the weirdness
> enters when considering whether it is ethical to train computer to learn
> and improve an algorithm until it is "perfectly ethical", if what it needs
> to do to get there is to fail several times before it can actually become
> perfect. How many failures should there be before it's not ethical to
> continue training the computer?
>
> I would say it's not ethical to do that, if it means for example
> surveilling a population with face recognition technology until it is able
> to perfectly identify a criminal from his or her doppelganger. There will
> always be the risk of accusing an innocent person, which is not ethical.
>
> Algorithms usually don't take into consideration context. I recall
> Rosenblatt's work on perceptrons were a way to create context by computers
> learning about contexts (by sensing). That actually might safer than
> constructing algorithms.
>
> You have to wonder what computers would be like now if Rosenblatt had been
> able to pursue his work unfettered by Minsky and others from MIT back then.
>
> Academic freedom must be protected. On that I hope we can agree!
>
> Kind regards,
>
> Annalisa
>
>
>
>
>
>
>
>
> ------------------------------
> *From:* xmca-l-bounces@mailman.ucsd.edu <xmca-l-bounces@mailman.ucsd.edu>
> on behalf of Edward Wall <ewall@umich.edu>
> *Sent:* Saturday, December 21, 2019 2:37 PM
> *To:* eXtended Mind, Culture, Activity <xmca-l@mailman.ucsd.edu>
> *Subject:* [Xmca-l] Re: The ethics of artificial intelligence, past
> present and future
>
>
> *  UNM-IT Warning:* This message was sent from outside of the LoboMail
> system. Do not click on links or open attachments unless you are sure the
> content is safe. (2.3)
> Annalisa
>
>      In my read when Dreyfus wrote the book you reference, he believed
> that ‘mind' was neither ‘material’ nor ‘mental’ On the other hand, I have
> often wondered if ‘minds' aren’t ‘material.'
>
> Ed Wall
>
> Imagination was given to man to compensate him for what he is not, and a
> sense of humor was provided to console him for what he is.
>
> On Dec 21, 2019, at  1:22 PM, Annalisa Aguilar <annalisa@unm.edu> wrote:
>
> Hello fellow and distant XMCArs,
>
> So today I saw this in the Intercept and thought I would share for your
> awareness, because of the recent developments that likely impact you,
> namely:
>
>    - the neoliberalization of higher academic learning
>    - the compromise of privacy and civil life in the US and other
>    countries
>    - the (apparently) hidden agenda of technology as it hard-wires biases
>    and control over women, minorities, and other vulnerable people to
>    reproduce past prejudices and power structures.
>
> In my thesis I discuss historical mental models of mind and how they
> inform technology design. During reading for my thesis I had always been
> bothered about the story of the AI Winter.
>
> Marvin Minsky, an "august" researcher from MIT labs of that period, had
> discredited Frank Rosenblatt's work on Perceptrons (which was reborn in the
> neural networks of the 1980's to early naughts). That act basically
> neutralized funding of legitimate research in AI and, through vicious
> academic politics, stymied anyone doing research even smelling like
> Perceptrons. Frank Rosenblatt died in 1971, likely feeling disgraced and
> ruined, never knowing the outcome of his lifework. It is a nightmare no
> academic would ever want.
>
> Thanks to Herbert Dreyfus, we know this story which is discussed in What
> Computers Still Can't Do
> https://mitpress.mit.edu/books/what-computers-still-cant-do
>
> Well, it ends up that Minksy has been allegedly tied up with Jeffery
> Epstein and his exploitation of young women.
>
> This has been recently reported in an article by Rodrigo Ochigame of
> Brazil, who was a student of Joichi Ito, who ran the MIT Media Lab. We know
> that Ito's projects were funded by none other than Epstein, and this reveal
> forced Ito's resignation. Read about it here:
> https://theintercept.com/2019/12/20/mit-ethical-ai-artificial-intelligence/
> <https://theintercept.com/2019/12/20/mit-ethical-ai-artificial-intelligence/?utm_source=The+Intercept+Newsletter&utm_campaign=0277d72712-EMAIL_CAMPAIGN_2019_12_21&utm_medium=email&utm_term=0_e00a5122d3-0277d72712-124483985>
>
> I have not completed reading the article, because I had to stop just to
> pass this on to the list, to share.
>
> One might say that computer technology is by its very nature going to
> reproduce power structures, but I would rather say that our mental models
> are not serving us to create those technology tools that we require to
> create an equitable society. How else can we free the tools from the power
> structures, if the only people who use them are those who perpetuate
> privilege and cheat, for example by thwarting academic freedom in its
> process? How can we develop equality in society if the tools we create come
> from inequitable motivations and interactions? Is it even possible?
>
> As I see it, the ethics at MIT Labs reveals concretely how the Cartesian
> model of mind, basically normalizes the mind of the privileged, and why
> only a holistic mental model provides safeguards against these biases that
> lead to these abuses. Models such as distributed cognition, CHAT, and
> similar constructs, intertwine the threads of thought to the body, to
> culture, history, tool-use, language, and society, because these models
> encapsulate how environment develops mind, which in turn develops
> environment and so on. Mind is not separate, in a certain sense, mind IS
> material and not disembodied. It is when mind is portrayed otherwise that
> the means of legitimizing abuse is given its nutrition to germinate without
> check.
>
> I feel an odd confirmation, as much as I am horrified to learn this new
> alleged connection of Minsky to Epstein, how the ways in which as a society
> we fool ourselves with these hyper-rational models which only reproduce
> abusive power structures.
>
> That is how it is done.
>
> It might also be a reminder to anyone who has been unethical how history
> has a way of revealing past deeds. Justice does come, albeit slowly.
>
> Kind regards as we near the end of 2019,
>
> Annalisa
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucsd.edu/pipermail/xmca-l/attachments/20191222/50fd9d89/attachment.html