[ PRIVACY Forum ] Script of my Monday national radio report on Meta monitoring of employees to train AI, and its broader implications to Big Tech and all of us

[Lauren Weinstein]

This is the script of my national network radio tech report last
Monday on the topic of Meta's monitoring of employees' computer
activities, using that data to train AI -- and discussion of the
broader implications of such projects. As always there may have been
minor wording variations from this script as I presented this report
live on air.

- - -

So yeah while this discussion is about Meta the same trend is
happening at other Big Tech firms where employees are increasingly
being subjected to mass layoffs or feel that they are being mistreated
by management, and the impact of this could be quite dangerous for
everyone who depends on services from these firms as the firms try to
use AI as an excuse for all manner of bad corporate behaviors.

But regarding Meta, it's useful to recall that Meta's CEO Mark
Zuckerberg has never been the poster boy for particularly admirable
corporate leadership. Of course before Meta was Meta it was Facebook,
and before that it was The Facebook and before that it was Facemash.
This takes us all the way back to when Mark started this entire sordid
set of events when, while a student at Harvard, he built a platform to
help guys avoid dating what he and his Harvard bros considered to be
unattractive women, sometimes using terms like "farm animals" for
their comparisons. Apparently he threw this project together after
being dumped by his then girlfriend.

This was all back around 2003. Facebook didn't change its name to Meta
until around five years ago or so, when Mark became convinced that
everyone would want to wear virtual reality headsets, working and
playing in his virtual reality "metaverse" all day, and threw
reportedly tens of billions of dollars at that project, some reports
say 80 billion or so. And while there are continuing to be amazing
applications for VR and Meta VR headsets, the idea that this was going
to be a dominant way for people to routinely interact was, what's that
technical term? -- oh yeah, crazy.

Of course over the years Meta/Facebook has been embroiled in a wide
variety of issues related to privacy and a long list of other
controversies. So now Mark has changed his mind again, and has decided
that he should throw most of Meta's resources at -- you guessed it --
AI. And the way he did this was for example, reportedly by reassigning
product managers and skilled engineers to be essentially AI trainers,
understandably considered to be work far below their skill levels, and
laying off other employees. Then Meta came up with the idea that is
somewhat akin to driving a stake through the heart of employee morale,
already apparently at historically low levels for the firm. And while
I don't know offhand who specifically originated this train wreck of a
concept, it seems like something Mark would have quickly and
enthusiastically endorsed.

So this is called MCI -- Model Capability Initiative. This was
launched just a couple of months ago, and collected employee
keystrokes, mouse movements, screen shots, etc. to be used for AI
training. Initially there was no opt-out. Then they apparently
introduced a limited half hour opt-out. As you MIGHT imagine this
didn't go over well with most employees. Even when using a company
provided computer as in this situation, employees routinely deal with
personally sensitive subjects including financial and medical
information and much more. And you can't blame employees for being
decidedly unenthusiastic about being forced to train AI systems that
they quite reasonably suspected Meta would like to use to lay them off
and replace them.

Well, this was all bad enough, but then very recently the fan really
got splattered when it was discovered that the data being collected
this way had not been properly secured internally, and was reportedly
accessible to pretty much anyone at the firm. Whammo. So about a week
ago the MCI program was suspended for some indefinite period -- maybe
forever, we don't know -- while "investigations" take place into how
this happened. This whole awful affair seems pretty much on-brand for
Zuckerberg, but again it would be a mistake to think that this kind of
situation could only occur at Meta.

Because across the technology sector we see firms increasingly
treating the employees who actually built these firms as expendable,
sometimes to be replaced with half-baked, dangerous
misinformation-laden AI, that drives customers and employees alike
utterly nuts.

But it does seem like Large Language Model AI is close to "jumping the
shark" as the saying goes, or perhaps already has, and is starting to
face some serious reckonings. That recent German court decision I
discussed recently, holding Google responsible for the content of
their AI Overviews -- that spout 10s of millions of wrong answers an
hour -- could be just the beginning.

If courts begin widely holding Big Tech AI firms responsible for their
AI-generated content -- search answer overviews, chatbots, and so on,
this very reasonable, common-sense approach could trigger massive
changes in the way these firms operate and might signal an end to the
abandon with which AI-created misinformation, sometimes dangerous and
harmful, is so widely spewed.

How these firms treat their own employees could be viewed as somewhat
akin to a canary in a coal mine regarding how these firms will treat
all of us. And judging from the mess at Meta in this respect, we
probably shouldn't expect a "customer is always right" approach from
Big Tech firms, that's for sure.

- - -

L

- - -
--Lauren--
Lauren Weinstein
lau...@vortex.com (https://www.vortex.com/lauren)
Lauren's Blog: https://lauren.vortex.com
Mastodon: https://mastodon.laurenweinstein.org/@lauren
Signal: By request on need to know basis
Founder: Network Neutrality Squad: https://www.nnsquad.org
PRIVACY Forum: https://www.vortex.com/privacy-info
Co-Founder: People For Internet Responsibility
_______________________________________________
privacy mailing list
https://lists.vortex.com/mailman/listinfo/privacy