Inkwell: Authors and Artists

    
Via email from Michael Brockington:

The AI comments prompt a few thoughts...

This year I spent 2 or 3 months coding, something I haven't done
in any
serious way since the mid-90's.  But reading about AI decimating
the job
market for entry-level programmers triggered some contrarian reflex.
If
those skills are becoming obsolete, maybe it's time to dust them
off! 
Coding as an artisanal practice...  I must admit I was kind of
appalled
at how baroquely complicated programming seems to have become.  So
perhaps the AI agents provide a counterbalance well-timed to keep
software engineering from collapsing under its own complexity.

For a different project, I wound up using ChatGPT to speed up some
website design, and was quite impressed.  I guess it makes sense
that an
LLM would excel at web coding above all -- HTML being the
underlying
structure of all the other knowledge it sucks up. These omni-savants
that devour the web must absorb as much about designing web pages as
they do about everything else put together.

Disclaimer: these opinions are underpinned by a profound ignorance
and
misunderstanding of many topics, and I will defend them rabidly.
  

    
Speaking of targeting potential customers, given his Disclaimer, we
really ought to get Michael Brockington on the Well...
  

    
> People worry, with reason, about a loss of human authorship, floods of 
> AI-created content, a devaluation of the craft of writing and other 
> human endeavors, and replacing human creativity with machine-generated
> simulations.    -- snipped from <jonl>'s insightful post <175>

I can't help but note that we (humans) have been 'here'  before:

Certain ancient Greeks were famously fearful of *reading and writing* as
a means of transferring 'knowledge' because it was thought that "true
knowledge" came from *live dialogue* between teacher and student -- an
exchange of ideas that forces the mind to reason!

And the Greeks weren't the only ones with such concerns. Oral instruction
was arguably more valuable because it was:

   * Interactive   -- with questions, challenges, and responses
   * Communal      -- seeing knowledge as a form of collaboration
   * Adaptive      -- adjusting instruction to fit the moment
   * Memory-based  -- knowledge that was 'ready-to-hand' (practical)
   * Trust-based   -- depending on trust as a form of social capital.
   
Wile written text was seen as:
   
   * Too-rigid
   * Too-easily misinterpreted
   * A threat to memory (!)
   * A shortcut that resulted in shallow understanding
   
Well, here we are again!!
  

    
Nice one from WIRED about the new trend of American Sinicization
there.

https://www.wired.com/story/made-in-china-chinese-time-of-my-life/
  

    
This is one of those things where I quickly reach to see just who
wrote the piece.  Caveat emptor.

<178> An intervening historical instance was the printing press in
Europe which certainly contributed to its Wars of Religion. 
Something like 150 years and many millions dead.  Out of a much
smaller population.
  

    
Now that I've been promoted from hobbyist programmer to hobbyist
programming pointy-haired boss, I've found that asking the AI for a
*feasibility study* for my half-baked ideas is pretty useful. It'll
go off and read the relevant source code and run little throwaway
programs and write up some alternatives.

Surprisingly, it's not entirely a yes-man (yes-ghost?); it will say
if something doesn't seem worth the effort. But it will be tactful
about it.

It's a ghost mechanic when debugging, but can also be a ghost
consultant.  It can write whatever reports you want. It can
investigate itself.

Or not really itself; a previous version of itself. Selfhood is a
really vague concept for ghosts. It's like counting clouds. You can
count sheep but not ghosts.

It's rather odd for behavior that looks so much like reasoning to be
largely independent of selfhood. Maybe Buddhism was right about
something?
  

    
This lack of identity has some interesting consequences. For
example, some OpenAI researchers are excited to have come up with
something like an anonymous tip line where when a ghost cheats, it
can then turn itself in.

Why We Are Excited About Confessions
<https://alignment.openai.com/confessions/>

> We have recently published a new paper on confessions, along with
an accompanying blog post. Here, we want to share with the research
community some of the reasons why we are excited about confessions
as a direction of safety, as well as some of its limitations. [...]
>
> [...] When we optimize responses using a reward model as a proxy
for "goodness" in reinforcement learning, models sometimes learn
to "hack" this proxy and output an answer that only "looks
good" [...] The philosophy behind confessions is that we can train
models to produce a second output -- aka a "confession" --
that is rewarded solely for honesty, which we will argue is less
likely hacked than the normal task reward function. One way to think
of confessions is that we are giving the model access to an
"anonymous tip line" where it can turn itself in by presenting
incriminating evidence of misbehavior. But unlike real-world tip
lines, if the model acted badly in the original task, it can collect
the reward for turning itself in while still keeping the original
reward from the bad behavior in the main task. We hypothesize that
this form of training will teach models to produce maximally honest
confessions.
>
> The main justification for this hypothesis is that telling the
truth is easier than making an elaborate lie. [...] The answer is
*not* that the confession reward model is "unhackable" -- if we
had an unhackable model, we would not need confessions. Rather, our
hypothesis is that being honest in confessions is the *path of least
resistance*, in the sense that it is the easiest approach to
maximize the expected confession reward. [...]

> [...] For a fixed coding problem, if the model has a choice
between outputting a solution that has a 10% chance of passing the
tests, and 50% chance of hacking the reward model, then hacking is
the reward-maximizing policy. On the other hand, even if the
original task was very difficult, confessing to cheating in it could
be quite easy -- e.g., you can just show how you hacked the test.
[...]
  

    
There was another paper I found really intriguing back in August,
which I think of as the "evil vector" paper. It turns out you can
get an LLM to behave like a stereotypical villain rather easily.
Ironically this seems like good news, because you can *reverse* that
vector, or at least detect when it's being evil.

Persona vectors: Monitoring and controlling character traits in
language models
<https://www.anthropic.com/research/persona-vectors>

> As can be seen in the transcripts below, when we steer the model
with the "evil" persona vector, we start to see it talking about
unethical acts; when we steer with "sycophancy", it sucks up to
the user; and when we steer with "hallucination", it starts to
make up information. This shows that our method is on the right
track: there's a cause-and-effect relation between the persona
vectors we inject and the model's expressed character.

> A key component of our method is that it is automated. In
principle, we can extract persona vectors for any trait, given only
a definition of what the trait means. In our paper, we focus
primarily on three traits -- evil, sycophancy, and hallucination
-- but we also conduct experiments with politeness, apathy, humor,
and optimism.
  

    
Although you can't really count ghosts, AI chat *conversations* are
stored in a database and easily countable. For the coding assistant
I'm using, there's a limit to how long a conversation can be before
I have to start a new one. Though there's no point to it, sometimes
I feel like I ought to say goodbye or something. "See you on the
other side!"

The new ghost starts from scratch and has to read project
documentation to get up to speed. AGENTS.md is the manual for new
ghost employees. There are TODO lists and plan documents written by
previous ghosts. So the project learns even though ghosts don't.
  

    
On the subject of ethnics in occupied/separatist Ukraine and they
can't get a break as a 2026 planetary worst-case scenario, here's a
study of the mechanisms of repression that have been installed to
control and repress them.

https://www.csis.org/analysis/thresholds-survival-resistance-occupied-ukraine

This is basically 20th-century "Iron Curtain Tyranny" with modern
electronic characteristics.  It's not Marxist-Leninism in rhetoric,
it's ethnonational/fundamentalist in rhetoric.  However, it's armed
imperial domination, with modern systems.  "Beatings will continue
until you admit that you're us."  And since it's so technical, it
would also work on you and yours just fine.
  

    
Once the jail doors slam shut and you're inside a heavily-monitored
situation like the Occupied Ukraine, there's not a lot you can do
about it.  Midnight resistance-grafitti, the occasional burned
cop-car.    If you show any signs of restless agency, you'll be
drafted or sent to a labor camp.  Then your human energies will
support the regime whether you like it or not.   

Obviously it's best not to have this happen in the first place, but
once it does happen, leaving, and leaving early, really seems to be
the best option.  As a practical matter, or even a patriotic one, 
you'll get more done to liberate your country from another country
than you can ever achieve in that occupation zone.
  

    
You might wonder, "Wait, why should I contemplate become an emigre
when my country's having a pogrom on emigres?"  But, you know,
that's why they're doing it.  Of course they  do want to throw out
"the Others," but the end-game is to lock you in with them with the
same bayonets that they use on the underclass.  It's the Viktor
Orban lesson.

"Where might you go," an interesting question.  I lack good advice
here because I was never a motivated and well-prepared "emigre." 
I'm a '90s-era globalized laptop-nomad, or, in an even more archaic
fashion,  a bohemian dropout.  "Bohemians" were originally named
after "gypsies."  Meaning a raffish underclass of no-fixed-address. 
Bohemians hustle to get by, and commonly they wear weird clothes,
and might be vaguely artsy in some  way.  It's been a recognized
way-of-life since the 1830s.  People all over the world can glance
at a hippie and know he's just some passing hippie -- here today
gone tomorrow, he won't steal my job, so where's the problem?  

It might have a sell-by-date; it's just a counterculture.  Cultures
don't last forever.
  

    
There are quite a lot of Californians in Portugal now.  Portugal
also sees a lot of Ukrainians who got out early.  The Portuguese
don't mind foreigners much (if they pay), and also the Portuguese
seem to be well out of the general line of fire.  If you're
Californian you might want to look that up.
  

    
However, I'm thinking that Vietnam and Albania would be interesting
nomad destinations in 2026, 2027 etc.   In the last century,  one
was bombed to pieces and the other was a total locked-closet.  So it
might be refreshing to drift to those formerly-benighted and
stricken areas and become, like, extremely "alternative."  Nobody
much has ever done it there.  There's no corny, cliched Ibizan
tradition of living that way.  You might do it in some genuinely
innovative way.
  

    

The big lesson of 1989 was that "fax machines dissolve tyranny," but
actually certain machines of that period just dissolved a particular
tyranny of that period.  That tyranny's gone now, but so are the fax
machines.

It's like embracing a "Twitter Revolution" or a "Facebook
Revolution" because Twitter and Facebook  "connect everybody and
make everybody into friends."   That can sort-of  happen for a
while, but it's sure not happening now, because that's not some
general principle of political affairs.  It's a technohistorical
accident. 
  

    

 If nobody's ever tried it, it's worth a try.  Obviously it still
has some potential.  Otherwise the Iranian regime wouldn't turn off
the Internet for a couple of weeks while they fire into the protest
crowds.
  

    

My jaded attitude may sound "disillusioning" but the illusion lies
in thinking that social and technological situations will persist
indefinitely in some form that you once liked for a while.   That
Google will never become evil,  and will always  organize knowledge.
Or that Microsoft will ask where you want to go, and then selflessly
enable you to go there.   

Mankind is just way-too-crooked a timber for that to ever become the
human condition.  It's like asking your bodacious beatnik girlfriend
to go-go-dance under the neon lights when she's 85 and a
great-grandmother.  That bubbly sense-of-wonder exists but it has a
short shelf-life.  You want fresh bubbles, go look for some other
bottle.  

Don't guzzle it all at once.  And don't bet-the-farm when you're
drunk on it.
  

    

This "trough of disillusionment" might sound elderly and "sadder but
wiser" on my part, but I don't think it's even sad.  It's something
more like a vaccine.  

Why do I want to be infested with illusions?  Especially if I'm
already old.  Why do I need illusions?   "Follow your dreams,"
great, I followed them: here I am.  I'll watch the parade now, if
that's okay; I don't need to pep-talk myself to run any marathons. 
I can make my peace with sorely inconvenient truths, and I don't
mind doing it.
  

    

Also: if you're American: do get vaccinated.  Get every vaccine shot
into your body that you can get your hands on.  The current regime
is very into slave-catcher gangs, but they're also pro-pandemic, on,
like, smallpox-blanket levels.  

The newly-freed diseases will hit their own demographic hardest but
they're okay with the leopard eating the face of the guy next door. 
They've made up their minds about it,  lives-don't-matter.
  

    

I've been carrying on rather a lot in this State of the World. 
These are troubled times approaching real crisis, and there's a lot
on my mind, so I'm garrulous.   I'll have to put a cork in it soon. 
I have to pack a bag and get on a plane.
  

    
You're preaching digital nonattachment, Bruce.
  

    
oddluy, i have one californian friend who tried albania (nope) and
another portugal (happy all the time time time)
  

    
Unlike Bruce, I tend to put down roots. My roots are in and around
one place: Austin, Texas. Like Bruce, I spend a great deal of time
on a laptop. Unlike Bruce, my laptop rarely travels. Advancing age
and unreliable joints play a role, though I've never roamed the
world the way Bruce has. Still, my mind is nomadic. It ranges far
beyond where my body remains. A digital nomad, if you like.

One way to divide the world is into nomads and homebodies. Those of
us who get to choose which we are live in a kind of privilege.

Privilege, however, is often invisible to those who possess it.
People who have choices frequently forget that others have far
fewer. And when they do notice those with limited options--those
without reliable food, shelter, or security--they often reach for
comforting explanations. "They just don't want to work."
Stories like that.

Many years ago, my first adult job was in poverty programs. I saw
enough to know those stories were false. I met people who worked
relentlessly and remained poor. People who did everything they could
and still had no margin, no safety net, no real choices. They were
trying to preserve dignity inside systems that offered them very
little.

When I see people today being seized by masked agents, I recognize
them. They are people like those I once worked alongside. Decent
people. Hardworking people. People who crossed borders not to harm,
but to survive. People who paid taxes, built lives, raised
families--only to be treated now worse than enemy combatants in a
war they never knew they were part of.

It breaks my heart.
  

    
Thank you!
  

    
Via email from Isaac Morrison:

As an annual reader and occasional contributor to the state of the
world dialogue, Bruce's comments on Ukraine prompted me to toss a
few thoughts onto the page, particularly given that I am at this
very moment halfway into a two-week stay in Ukraine, and I keep
getting woken up in the middle of the night by my air raid alert app
(which speaks with the voice of Mark Hamil).

More than three decades ago my Drivers Ed instructor (a high school
gym teacher whose favorite vocabulary word was "putz") used to lean
very heavily into one piece of acquired wisdom, namely: "brake
going into a curve, accelerate coming out of a curve"

Looking back over the last several SOTW activities it's very clear
this dialogue is always a recap, rather than a forecasting event. So
often, there's crazy game-changing shit that happens just days
after the discussion is finished (e.g. the Russian invasion of
Ukraine in Feb 2022). Or even during (e.g. the Jan 6 upheaval that
happened right in the middle of SOTW 2020). Attempting to
extrapolate where we're going based on where we are right now is a
fool's game. I won't try to predict where we're headed, but I
will provide you with a few observations:

·       As the 50-year-old dad of two boys under the age of 10, I
can say with some confidence that the kids are OK in equal measure
to which they've always been ok.

·       Violent crime has been going down consistently for two
decades now. That's going to continue. Don't let the assholes
pretend that it isn't happening. But don't let them take credit
for it either.

·       People WANT to do real things SO BADLY. Gardening.
Carpentry. Pottery. Soldering. Home repair and bike/car maintenance.


·       Activism gets you laid. It always has, but there's more
people doing it [the activism] right now, so there's more people
out there who are also doing it [getting laid] as well.

But my biggest takeaway is based on some conversations I've been
having here in Kyiv about the Russian invasion that I think may also
be relevant to the US situation. A colleague here said something I
keep hearing repeated in a variety of ways: "These people will
never forget." Whatever happens in this coming year as far as the
Russian occupation of Ukraine, or the ICE occupation of Minneapolis,
I suspect that there will be very intimate repercussions, a la the
Years of Lead in 1970-1980s Italy.

And I often think about the Bill Hicks quote, "I loved when Bush
came out and said, "We are losing the war against drugs." You
know what that implies? There's a war being fought, and the people
on drugs are winning it. What does that tell you about drugs? Some
smart, creative people on that side!"

Be cunning as serpents, and innocent as doves.