Putting the Stargate AI Hype into Perspective
You’ll do yourself and the world a tremendous service if every time you see or hear the term “AI” you substitute the phrase “computer model.”
This is because so-called artificial intelligence (AI) is a computer model — or more specifically, a collection of many such models. When you hear “computer model” instead of “AI,” you’re much less apt to become excited or nervous, and when you speak the substitution it becomes much less easy to boast or to have the boast taken overly seriously.
With that in mind, let’s look at the buzz around the government’s new Stargate Project. Not the old Star Gate Project from the 1990s that was run by the CIA with a particular concern for what the Chinese and Russians were up to. The new Stargate (sans space) is about computer models, with a particular concern for what the Chinese and Russians are up to.
Three big things and one small one were announced for Stargate, not counting new jobs. (There isn’t a government program created that doesn’t promise new jobs or a soaring economy, so this can be ignored.)
Eye See You
The first big thing is ubiquitous constant surveillance. Oracle’s Larry Ellison, who’s in on Stargate, is keen on this. He said that “citizens will be on their best behavior because we’re constantly reporting and recording everything that’s going on.” He also said computer modeling finagling these streams of data will make “unimpeachable” judgments.
Which is false.
The National Security Agency, and therefore other intelligence-connected entities, have been spying on, well, everybody since the moral panic over terrorism. Ellison is right that, if he is allowed, computers (inside and outside government) will record everything, just as they now track you everywhere you go, everybody you contact, how long and where the conversations take place, and much more if you carry a smartphone.
Ellison wants to expand this surveillance on the idea that utopia can at last be reached once computer models are put in charge of processing the recorded images and sounds. Yet if you’ve ever watched an American sporting event with an official timeout, referees studying a replay from every conceivable angle and coming to a decision that half the crowd disbelieves, you’ll understand why Ellison’s sources will be highly impeachable.
Yet lest we be too sanguine, let us recall Cardinal Richelieu, who said, “If you give me six lines written by the hand of the most honest of men, I will find something in them which will hang him.” AI can, if allowed, automate this process.
Boastful Promontory
The second thing promised is computer general modeling — a term which doesn’t get the juices flowing. It sounds a lot slicker without the substitution: Artificial General Intelligence, or so-called strong AI. These are computer models that, somehow, become “self-aware” and possess intellects and wills as we do. (Anyone who’s ever seen any of the Terminator movies will be intimately familiar with this concept.)
This claim is a bluff. This is a deep and complex topic, and we can’t solve it here except to note that because nobody — and I mean nobody — knows how our own intellects and wills are created, as they are immaterial, it is pure bravado to say we can create computer code to mimic these rational structures. If we don’t know how we work, which we don’t, we certainly can’t say how mere computer models can “come alive.”
Yet even those who know better can’t shake the feeling of dread over this topic. Even the Vatican’s new attempt to calm fears over AI, Antiqua et Nova, the “Note on the Relationship Between Artificial Intelligence and Human Intelligence,” contains this line: “AI cannot currently replicate moral discernment or the ability to establish authentic relationships.”
Cannot currently? Try cannot ever. AI is only a model, and all models only say what they are told to say. Computer models will never replicate discernment of any kind, because there is no mind in the code that can discern.
The Flowing Red Stuff
The third thing, and the one of most interest, is the claim about the “cure for cancer.” Most of the news reports read like marketing hype:
Ellison explained that cancer detection could begin with small fragments of tumours present in the bloodstream. AI could assist in identifying these fragments early through a simple blood test, followed by gene sequencing of the tumour. Based on this sequencing, a vaccine could be designed to target the cancer. He emphasised that this approach could allow for early detection and rapid vaccine production, with the latter being achievable within two days.
Blood tests are not easy, as the spectacular failure (would con be too strong a word?) of Theranos proved.
Computer models can help here in a secondary way — by which I mean that the hard work of correlating chemical signatures in blood with the presence of tumors in the body and distinguishing these hideous complexities from other signals first has to be done, and largely by hand. Computer models can aid in this, as they can in any statistical analysis. But producing quality data comes first, which is outside “AI’s” direct purview.
In any case, nobody is riled over blood tests. The real controversy lies in the boast about AI helping create personalized mRNA vaccines — all of which, alas, we are forced to view through the lens of the COVID panic.
The Wrong Messenger
The lying, obfuscation, fear, blustering, bullying, outright tyranny, lousy science blunting good science, hubris, over-certainty, and rank idiocy of that panic makes it almost impossible to think straight about this topic.
At the time, CDC director Rochelle Walensky marched in front of cameras and told mighty lies, such as if you got the mRNA shot you couldn’t get sick and couldn’t pass on the bug. Experts insisted that there was no chance of any adverse event (a first for a medicine). Elites backed these lies, exaggerated the risks, and shamed those who feared getting this new form of drug. Not to mention all the people who lost jobs or were forced to get vaccine passports.
It was only natural that a backlash against the shot arose. Any discussion of these treatments now is still viewed with great suspicion. But let’s burn through the fog.
There are two aspects of interest: the utility of the vaccines themselves and the role computer models played in creating them.
COVID has already proven that mRNA vaccines are useful but imperfect, and do indeed carry risks. They were not newly invented for COVID, and had already been studied for some years — even in cancer treatments.
One thing has already become clear: These are not panaceas. The 2023 Nature Reviews Cancer paper “mRNA-based cancer therapeutics” reports that one trial for prostate cancer was canceled in 2017 because the shots weren’t working. Another trial for melanoma did show increased immune system activity, but with “influenza-like adverse events.”
Additionally, mRNA therapies might not be able to target all cancer types, because the antigens in some cancer types are also present in normal cells. Nevertheless, it is possible to sequence some kinds of cancer cells and fashion tumor neoantigen vaccines to target these sequences. These vaccines have to be delivered by some mechanism, such as coating them with lipids (as with COVID), which can build up in cells and cause adverse events.
For instance, one trial of an interleukin-12 mRNA vaccine was “hampered by toxic side effects (for example, fatigue, dyspnoea, stomatitis, acidosis and gastrointestinal haemorrhage) and severe adverse events, including death” — the mother of all side effects.
Therefore, it’s not impossible that mRNA treatments will bring benefits, but the difficulty in making them is the same as with blood tests: The brutal benchwork in discovering cause and effect has to be done first. The mRNA cancer therapies are not ready for prime time — and for some cancers may never be. Computer models can and must be helpmeets, but they will not create discoveries. The best thing another different Nature paper could say was that the “integration of AI-driven methodologies into the drug development pipeline has already heralded subtle yet meaningful enhancements” in the process. Good news, but nothing to grow short of breath over.
The real problem is loss of privacy: submit to these blood tests, and your DNA belongs to a company to do with as it pleases despite any promises made about secrecy. Do you trust them? As with cellphones, many people already do, blithely sending DNA samples to companies to gain insight into their genealogies.
The Arrival of Abby Normal
We at last come to the one small thing promised by those involved with Stargate: Sam Altman of OpenAI gushed that “we will see diseases get cured at an unprecedented rate.” This is a largely empty promise. But he also sees computer models bringing him immortality. This will happen, transhumanists maintain, in two ways: by using computer models to hack the genetic code using computer models, and, if that doesn’t work, by “uploading” one’s consciousness into one.
I say this is a small thing because, as with strong AI, it’s a bluff. If you don’t know how minds work, which no one does, then you will not be able to create a computer that has one — and there is no chance of “uploading” your own mind into that which does not exist.
I have no proof the genetic code cannot be hacked, but I do know the overcertainty here is vast — vast! — and that if men like Denis Noble are right (and I think he surely is), then genes alone are far from the whole story. All talk about mechanistic immortality is wishful thinking.
There are many promising aspects of AI, such as automation of mundane tasks, but there are also important limits and dangers, like fake videos and tyrannical surveillance — not to mention people forgetting that AI is just a collection of computer models and putting too much trust in them.
William M. Briggs is a contributor to The Stream, the author of Uncertainty and maintains an active and lively blog at wmbriggs.com. He earned his Ph.D. from Cornell University in statistics and studies the philosophy of science, the use and misuses of uncertainty, the corruption of science and the uselessness of most predictions. He began life as a cryptologist for the Air Force, slipped into weather and climate forecasting, and matured into an epistemologist.