Reading:

Why machines will never rule the world

J. Landgrebe, B. Smith

#AI #unplugai #LLM #AIhype

If you want a career in #AI now, look to train models with minimal #data.
The web is forever poisoned now with #disinformation.

#unplugai #llm

#LLM is not the pinnacle but actually the nadir of #bigtech

#unplugai #TeamHuman

A financial #crash caused by “woke” #AI is a non-negligible risk for the #Biden camp.

Difficult to navigate.

Best scenario would be a public scandal sooner than later. Then, Biden could send in FTC, SEC etc to “cool” the industry like they did in #crypto after last years scandals.

This way Biden will avoid to openly attack #BigTech but decrease the risk of a blow-out during the election campaign.

#unplugai #TeamHuman

There is nothing more cute than #AI Fanboys using #GPT4 API to plan household chores or hobbyist programming and touting their “business case”.

#FinancialCrash #AlgoTrading #unplugai

#Emad of #Stability.AI boasts that #Hedgefonds are interested.

Prediction: #LLM based #AI will be outlawed after it causes the next global #financial #Crash.

Hedgefonds connecting #GPT4 via API to #highfrequency #trading algos will result in something like the #FlashCrash we saw in the past. Let’s see if the #SEC gets ahead of this train crash.

#unplugai

The #AI Bubble feels like the #Web3 Bubble exactly 24 months ago.

At the time #takeoff was imminent, about to change everything. People were quitting jobs to trade #coins, create #NFTs etc etc.

And the different #crypto currencies were pushing each other forward with ever bigger announcements of billion dollar investment deals.

Oh, and there was even a use case: Make #banking obsolete. Make the #corporation obsolete. Make #lawyers obsolete…

Sounds familiar?

#unplugai

If #GPT4 were a chemical it would be banned.

It might surprise to hear that very few #chemicals like mercury, lead and DDT are banned based on proof of harm beyond doubt (due to terrible tragedies that occurred).

It’s usually sufficient to give a high risk of harm based on strong circumstantial evidence in animal tests.

By that measure the suspicion that #GPT4 can be used to fool humans during decision making would be enough to ban it and all equivalent #LLM s.

#unplugai

#AIDoomers are afraid of #AI b/c they are certain it will become a #Superoptimizer or #Superintelligence , engineering away all #Life on #Earth - they speak from experience b/c #HomoSapiens is doing just that, using the superintelligence of global #corporations

So, #LLM AI is the moment another generation becomes afraid of its own species, in the mirror of technology.

Like their grandparents in front of #Auschwitz and #Hiroshima. And their peers confronted w/ #ClimateChange.

#unplugai

And this analogy is of course also a good indication , that #coding #copilot will remain the only useful application of #LLM , apart from writing boilerplate , bullshit, and propaganda.

#unplugai

So, what by this analogy what #LLM s show us :

Not that humans are limited but that the digital civilization is limited, chasing its own tail for two generations at least.

#BrunoLatour #unplugai

Reading Unger & Smolin (2015) opens up an interesting analogy:

In a mature, sufficiently cool universe, laws of generation (aka laws of nature) emerge from sufficiently stable recurrent causation.

In a sufficiently mature computing network, rules of generation emerge from feeding output back as input.

That’s a succinct definition of #autoregressive #LLM :

A model of a #System feeding on its own output.

#unplugai

Reminder why we must #unplugai :

“ Taking instructions that aren’t from individuals means ceding the power we hold as individuals, castrating yourself in service to the instructions you cannot hope to change.”

https://desystemize.substack.com/p/ai-the-idiot-ant-queen

#Idiocy is globalized.

It doesn’t burn itself between five frat boys like it did in the old days.

Today it’s burning itself out at the 100 bn dollar scale, in front of everyone with internet access.

#unplugai

#AI based on current GPU hardware, and #deeplearning will with 100% not achieve
- #sentience
- #consciousness
- runaway learning towards #superintelligence

Instead, it will forever depend on reasonable (human or other biological , maybe #corvids?) input to acquire new skills.

#Chatbots are the new #FSD , #selfdriving cars

#unplugai if you want to stop embarrassing yourself

And of course that s exactly the social pattern we have to come to know from a decade of #Blockchain

#unplugai

Nobody understands what #Deeplearning does precisely because nobody is able to critique their own or somebody else’s #theory of what is going on. It just words that could just as well be generated by a #LLM

#Understanding is #criticizing - and thus there continues to be no noticeable #Understanding in #DL

#unplugai

@ceperez

Here is a critique of your definition of #Understanding .

https://desystemize.substack.com/p/representation-and-uncertainty

#unplugai

The guy who tells us “#AI that I build might wipe out #humanity” has been suffering from a #GodComplex for most of his life: holy moly! This is worse than I thought.

When #AI and #Blockchain crash and burn together whom will we call for help?

#SamAltman #OpenAI

On the upside: Now #ClimateChange is back as the uncontested Number One Priority.

#unplugai

#OpenAi is just another corporate trickster out there to rip you off. They ship unsafe product, they have a major privacy breach but hide it, they pull research access destroying PhD projects - same old , same old.

OpenAI is a corporate #rogue of the sort we have come to know too well.

Good, anything else would have been scary.

#unplugai