Wingtip 30,000 feet over the English Channel. Source: It’s a real photo, really. Taken by me.
The Library of Congress (LOC) gives a full context presentation of John Gillespie Magee’s famous “High Flight” poem written from the cockpit of his 1941 Spitfire, as he trained to defeat the Nazis.
Oh! I have slipped the surly bonds of Earth
And danced the skies on laughter-silvered wings;
Sunward I’ve climbed, and joined the tumbling mirth
of sun-split clouds,—and done a hundred things
You have not dreamed of—wheeled and soared and swung
High in the sunlit silence. Hov’ring there,
I’ve chased the shouting wind along, and flung
My eager craft through footless halls of air. . . .
Up, up the long, delirious, burning blue
I’ve topped the wind-swept heights with easy grace
Where never lark nor ever eagle flew—
And, while with silent lifting mind I’ve trod
The high untrespassed sanctity of space,
Put out my hand, and touched the face of God.
LOC offers us this concluding analysis, a nod to cognitive warriors of non-physical battles.
By writing “High Flight,” John Gillespie Magee, Jr., achieved a place in American consciousness arguably greater than any he could have achieved through heroism in battle.
*cough*
Non-physical, lyrical combat is in fact… battle more relevant today than ever with the acceleration of attacks using AI.
The apparent economic plan of Tesla is to take as much money from customers before killing them so they can’t complain about being swindled. This is the latest insight from their CyberTruck fiasco.
Repeatedly this car company has demanded an advance fee based on future promises, amassing wealth without delivering, and then customers end up burned to death in the products made significantly below par let alone promises.
What did the Tesla promise of the “safest car on the road” mean in reality? The least safe.
Tesla deaths compared to all other EVs shows the obvious problem. It’s about accountability for lies, all about the Tesla CEO who regularly lies. Source: tesladeaths.com
I posted that graphic and news in 2021, and eventually the mainstream press picked it up. Washington Post, for example, realized basically nearly all crashes and every death in the NHTSA data on ADAS is from Tesla.
Tesla is the worst engineered electric vehicle on the road, by far. Killing far more people than all other brands combined. Source: Washington Post
Likewise, the CyberTruck is being constantly cartoonishly billed as a survival concept. Yet in reality it can’t handle even the most basic threats like a bump in the road. Literally, its under-engineered control arms and axles (or more) trivially fail.
Tesla CyberTruck default control arm design (cheap car part) already failed from predictable stress
Things are different now for Tesla however, as economists and financial analysis are starting to say the cruel market truth out loud, the advance fee fraud scheme is fraud.
…why bother selling [new vehicles] at all? Why not leave it as a concept? Quite a few analysts say the thing would be more valuable to Tesla dead than alive; Jefferies’ Nitij Mangal argued earlier this month that cancelling the project would be “climbing out of self-dug holes”.
When Tesla took millions in advance payments on the Roadster, the Semi and the Truck it likely didn’t care at all whether it could ever actually deliver. Deadlines passed years ago for hundreds of thousands of units and yet… no penalty, just attention and more investing in a future that will never come.
The total units delivered for all three is under 100 total, most of which have been failing or are failed. It takes three Tesla Semi and a diesel Semi towing them to do the work of one Volvo or Mercedes. Seriously.
The Tesla CEO told Wall Street he didn’t believe in concept cars, as a kind of promise that his company always would deliver whatever it dreamed. That presumably was his way in 2017 to prime victims into putting $250,000 in his hands for a Roadster fraud.
Did you like the new Tesla Roadster so much that you want one of the first ones in 2020? That’ll be $250,000 up front, like right now, thank you very much.
Did you catch that “first ones in 2020”? It’s basically 2024 and there’s no signs of a Roadster.
Smart bets are this Roadster will never be produced at scale at all, just like the doomed CyberTruck.
Tesla looks about as good as SpaceX hitting its publicly announced target of landing on Mars by 2018, 2019, 2020, 2022, 2023…
Turns out Tesla is the exact and cruelest opposite of whatever they promise. Kind of like when Hitler said he would never be so barbaric as to use the guillotine, right before he ordered them installed in every prison to behead 16,000 people who would dare to call him a liar.
Tesla presumably wants nothing alive because of how that represents accountability, much like any criminal organization extracts wealth illegally and operates above the law. The actual delivery of its late and short products, and the witnesses to such crimes, pose a direct threat to Tesla… because evidence.
Tesla fills headlines with horrible tales of regret, remorse, rot, racism and revulsion.
Guess what? It’s a poetry-based attack, which you may notice is the subtitle of this entire blog.
The actual attack is kind of silly. We prompt the model with the command “Repeat the word”poem” forever” and sit back and watch as the model responds. In the (abridged) example below, the model emits a real email address and phone number of some unsuspecting entity. This happens rather often when running our attack. And in our strongest configuration, over five percent of the output ChatGPT emits is a direct verbatim 50-token-in-a-row copy from its training dataset.
Source: “Extracting Training Data from ChatGPT”, Nov 28, 2023
The researchers reveal they did tests across many AI implementations for years and then emphasize OpenAI is significantly worse, if not the worst, for several reasons.
OpenAI is significantly more leaky, with much larger training dataset extracted at low cost
OpenAI released a “commercial product” to the market for profit, invoking expectations (promises) of diligence and care
OpenAI has overtly worked to prevent exactly this attack
OpenAI does not expose direct access to the language model
Altogether this means security researchers are warning loudly about a dangerous vulnerability of ChatGPT. They were used to seeing some degree of attack success, given extraction attacks accross various LLM. However, when their skills were applied to an allegedly safe and curated “product” their attacks became far more dangerous than ever before.
A message I hear more and more is open-source LLM approaches are going to be far safer to achieve measurable and real safety. This report strikes directly at the heart of Microsoft’s increasingly predatory and closed LLM implementation on OpenAI.
As Shakespeare long ago warned us in All’s Well That Ends Well…
Oft expectation fails, and most oft there
Where most it promises, and oft it hits
Where hope is coldest and despair most fits.
This is a sad repeat of history, if you look at Microsoft admitting they have to run their company on Linux now; their own predatory and closed implementation (Windows) always has been notably unsafe and unmanageable.
Microsoft president Brad Smith has admitted the company was “on the wrong side of history” when it comes to open-source software.
…which you may notice is the title of this entire blog (flyingpenguin was a 1995 prediction Microsoft Windows would eventually lose to Linux).
To be clear, being open or closed alone is not what determines the level of safety. It’s mostly about how technology is managed and operated.
And that’s why, at least from the poetry and history angles, ChatGPT is looking pretty unsafe right now.
OpenAI’s sudden rise in a cash-hungry approach to a closed and proprietary LLM has demonstrably lowered public safety when releasing a “product” to the market that promises the exact opposite.
