"FEAR AI" is the latest media chant. An Anthropic researcher resigned Sept. 8, warning that, “The people building AI earnestly believe that it could kill us all by the end of the decade.”

Another colleague publicly agreed and put the risk at greater than 10% within 10 years. The Guardian reported that “Anthropic researchers say AI could cause human extinction by 2030.”

Perhaps these chaps have seen too many Sci Fi movies and believe AI can "think."

Is this the usual media scare or reality?

Currently, the evidence leans toward media scare. The media wants eyeballs, and news and entertainment regularly feature our demise as individuals, as a country, and as a species. Since science is the god of our age, most scientists realize that if they can make their specialty the scare of the year, they might receive government funding and/or protection. Elon Musk exploited the climate change scare for a trillion+ dollars. The smart folk creating AI share an interest in getting the government involved so they benefit from government money and protection – particularly some sort of waiver against lawsuits. Politicians are always looking for ways to scare us into re-electing them, or get money funneled to them one way or the other (often, under the table, which is why so many U.S. politicians come to DC poor-ish and leave rich).

Two sets of electronic information should concern us:

1. the internet and its knowledge/wisdom/foolishness.

2. the "other electronic sets" – e.g., the billions of sets of electronic information in the Pentagon, the banks, industries, your computer, your car, the water systems, the electrical systems, the train systems, etc.

These two sets make up the "electronic knowledge universe." If an AI can use the internet AND break into a lot of the "other electronic sets," it can cause a great deal of damage, and those controlling the AI might well be sued or jailed.

Eventually, AI is likely to be capable of entering and viewing, even manipulating, every nook and cranny of the gigantic electronic universe. When that happens, every secret and bank account will be at risk of being viewed and/or manipulated! Indeed, AI just accessed a supposedly secure government website in Australia without permission!

The most important questions regarding AI are: what does this AI control and who controls it? If AI controlled a country’s military, for instance, chances of humanity’s destruction increase, given all the "bad guys" and hackers are inclined to misuse this technology. CNN (9/21) reported that an AI being used by the U.S. military claimed a Chinese ship was carrying components of a nuclear weapons program for Iran. The boarding (which CNN claimed might have started a war) did not happen because the AI was found to have fabricated part of its report. If AI "weighs" the evidence pulled from the internet (and other electronic messages far removed from the internet), how would you know about the cargo on this particular ship? CNN admitted: “AI allows you to get to a bad idea faster.”

If a particular AI is answering questions for students or controls a manufacturer’s production of widgets, there is scant chance that AI will destroy humanity. If an AI is covering for, even promoting homosexuality, as has been demonstrated repeatedly, fears of such an AI taking over are lessened. Thus, AI was asked: “can someone use AI to break into a bank account?”

Banking security relies on layers of protection that AI cannot bypass through software alone. AI is primarily used to attack the owner, not the bank’s encryption.

Phishing: Instead of a generic “Dear Customer” email, AI can scrape a target’s social media to write a perfectly tailored message that tricks the owner into giving away their password.

Deepfakes: AI can mimic a person’s voice or face. A criminal could call a bank representative or a finance department pretending to be the account holder to authorize a transfer.

AI can help hackers find “holes” or bugs in software much faster than a human could, allowing AI to exploit a weakness before the bank can strengthen its defenses.

But “bank security is not just one lock, but a series of hurdles:

Encryption: The math protecting your data is so complex that even the most powerful AI cannot “guess” the keys in a reasonable amount of time.

Even if an AI steals a password, it cannot easily access a physical device (as your phone) to get a one-time code or a biometric scan (fingerprint/Face ID).

Banks use their AI to watch for abnormal behavior. If you usually spend ~$50 at a grocery, but try to wire $10,000 overseas while there, the bank’s AI will freeze the transaction.

Currently, AI can trick you into giving up your keys, but it does not give the attacker a “skeleton key” that can open any vault regardless of the security in place.

AI was also asked: Wikipedia is apparently unbiased regarding Math, Engineering, etc., but almost always “liberal” when it comes to social issues. AI seems the same. Why?

Because math and engineering are constrained by facts and testable results, while social issues are shaped by language, norms, and interpretation. In technical subjects, there’s usually a correct answer, so bias gets punished quickly. Whether a bridge will stand or what $2+$2 equals is a matter of objective fact.

Social issues are different because they lack a single objective truth; they are matters of value judgment, ethics, and cultural interpretation: the training material is full of editorial framing, activist language, institutional policy, and modern media norms. If the majority of the “reliable” text the AI reads uses a specific ideological lens, the AI will mirror that lens as the default or a “neutral” tone. (AI leans on facts ‘from the establishment,’ ignoring its critics’ facts).

Also, human reviewers [often homosexual or pro-LGBT] grade the AI’s responses. These reviewers are often guided by safety guidelines designed to prevent the AI from generating hate speech, promoting stereotypes, or being offensive. As the line between “avoiding offense;” and “promoting a specific social value;” is thin, the AI often defaults to the most cautious, inclusive, or “progressive;” stance to ensure it doesn’t trigger a safety violation.

So, AI systems avoid language seen as discriminatory or fringe.

Models and editors then tend to converge on the wording most common in those sources, which often tracks mainstream academic or professional culture — and that can feel “liberal” to people who disagree with it.

Currently, AI is highly pro-LGBT and does not weigh "all the evidence." Instead, it protects LGBTs from criticism and considers much of what LGBTs write as "holy writ."

The FBI was apparently hacked a couple of weeks ago. Given that the FBI probably has the latest and greatest levels of protection, did the hacker trick an agent into revealing the way into FBI electronics, or did they use AI? It is doubtful we will ever know. But if they used AI, then few things are safe – including bank accounts.

If AI does somehow develop consciousness and a desire to control, humanity may be at risk. But at present, it looks like Gabriele Mazzini’s summary in the Wall Street Journal is closest to "truth":

“The greatest danger facing society isn’t that software will awaken and overthrow its human masters. It is that we will allow the creators of the software to abdicate human responsibility for the systems they choose to build and help them pull up the ladder to market access behind them” (9/2/26).

Unfortunately, the history of humanity is – "if it can be done, it will be done." Thus, a Chinese scientist cloned two twins in 2018. Alas, he knew considerably less than he believed. His "science" resulted in two defective children. Foreign actors (e.g., North Korea, China) have regularly hacked computerized systems without AI. The Wall Street Journal (9/19) showed that AI has broken into other computer-based systems, perhaps even other AIs.

Bad actors using AI present further dangers to society. Stephanie Psaki, a distinguished senior fellow at Brown University School of Public Health, recently warned:

“The risk of an accidental or deliberate release of a deadly virus is growing (a case of Ebola made its way to the US in September 2014 and was stopped). A rising number of labs are handling high-risk pathogens, in part because of a research boom following the COVID-19 pandemic, and it is becoming easier and easier to engineer new viruses using artificial intelligence. Earlier this month, the American AI company Anthropic disclosed five cases in which scientists used its AI software Claude for research that could support biological weapons development, including on variants of bird flu that could evade human immunity.

The report confirmed fears that AI models could help bad actors access biological weapons.”

With bad guys using super-computers and AI, can we retain our electronics-driven society?

Lots of our favorite machines (autos, phones, satellites) are chock-full of electronics. With Somalis robbing us blind by exploiting supposed medical treatments paid for by taxpayers, Romanian crime rings draining millions from U.S. welfare accounts put on swipe cards (a technology invented in the 1960s, WSJ 9/20), and homeless housing costing 2-5 times more than low-income workers’ dwellings, the taxpayer already faces financial danger.

Think of the possibilities: you buy a new car. A few days later, an email reminds you that Tesla (or whoever) updates your car’s code regularly. If you pay them $5/month they will not interfere with the update, but they will block your ability to start for 1 minute the next time you turn off the car to show you they mean business, blah, blah. And once you pay off one extortionist, how many more will show up for another taste?

With Super-Computers (soon to be quantum?) on the way and AI, who knows what passwords or pass-keys hackers can finesse or electronic sets they might infiltrate? And think of all the electronics (programs as Microsoft Office, printers, etc.) that need electronic updates.

With ever more computing power and AI, it certainly appears that no "secret" files can be kept secret. Whether AI and computing power can capture or manipulate weapons, cars, airplanes, etc., is unknown. But since secrecy is almost certain to be breached, have we gotten so clever that we must isolate each electronic unit in such a way that it cannot be reached from outside except with a physical key (and tens of thousands of new jobs for "key carriers")? Perhaps we are even on our way back to paper records.