Confronting the ‘Artificial Intelligence’ Backward-Looking Bureaucracy Trap

Former Secretary of State and White House national security administrator Henry Kissinger’s last service for his country was to warn his fellow Americans that substituting artificial intelligence for real human intelligence was “simply a mad race” toward “catastrophe.” 

Kissinger joined former Google CEO Eric Schmidt and MIT Dean Daniel Huttenlocher to write a formal critique of artificial intelligence by comparing it to the 1455 Gutenberg Bible. “Whereas the printing press caused a profusion of modern human thought,” they argued, AI would frustrate it by “creat[ing] a gap between human knowledge and human understanding.”

According to Kissinger, Schmidt, and Huttenlocher, the latest technology is called “generative pre-trained transformer” (GPT) artificial intelligence. OpenAI’s ChatGPT version “is now able to converse with humans“; its goal is to “redefine human knowledge, accelerate changes in the fabric of our reality, and reorganize politics and society.”

What is that new reality? Here is how the Kissinger team explained AI’s “change”:

Sophisticated AI methods produce results without explaining why or how their process works. The GPT computer is prompted by a query from a human. The learning machine answers in literate text within seconds. It is able to do so because it has pregenerated representations of the vast data on which it was trained. Because the process by which it created those representations was developed by machine learning that reflects patterns and connections across vast amounts of text, the precise sources and reasons for any one representation’s particular features remain unknown. 

They reported that AI “complexity … doubl[es] every few months” with “capabilities that remain undisclosed even to their inventors.” Inventors proceed “without understanding their origin or destination.” As far as results:

ChatGPT’s answers, statements and observations appear without an explanation of where they came from and without an identifiable author. On its face, ChatGPT has no discernible motive or bias. Its outputs are complex, but its work is astonishingly rapid: In a matter of seconds, it can produce answers that coherently explain a high-level topic. They are not simply copied from the text in the computer’s memory. They are generated anew by a process that humans are unable to replicate. 

The team presented questions to the machine that were answered wrongly, distorted, or even changed over time. Yet, answers come in seconds and seem so authoritative. Students already use ChatGPT to cheat on exams. It can be used to distort advertising to look as though it is from friends. It can be used to distort material from public officials and candidates for political office. How about deciding whether to go to war based on such information? (RELATED: Biden Administration Meddles in Technology It Doesn’t Understand)

Over the long run, the analysts cautioned:

Leadership [of artificial intelligence] is likely to concentrate in the hands of the fewer people and institutions who control access to the limited number of machines capable of high-quality synthesis of reality. Because of the enormous cost of their processing power, the most effective machines within society may stay in the hands of a small subgroup domestically and in the control of a few superpowers internationally.

But to this former director of the U.S. Office of Personnel Management, this conclusion merely gets to the surface of the real problem, which is our own superpower government. Its increasingly bureaucratic dominance of American domestic social life is only exceeded by popular dissatisfaction with the results. With the natural bias of government data toward the past, AI would make national government more and more outdated even if it worked as intended, since AI itself looks to past data. Even if the direct defects of AI itself can be minimized, the fundamental problem comes from government itself, which is the only institution that can force the results to be implemented.

Government bureaucracy developed historically by measuring crops as a way to bring the wealth of agricultural Mesopotamia under state control. That primitive version first created symbols to count grain, and similar but evolving information-like tools have been essential to its power ever since. And since such tools can only measure the past, the past has generally controlled the future. That is why it took so long to develop efficient markets, and why the rise of general prosperity did not take place until 10,000 years later in decentralized Europe rather than in the bureaucratic empires.

Artificial intelligence vastly exaggerates today’s more centralized, cumbersome, bureaucratic national governments’ dependence on the past. Modern AI is the past on steroids. But social progress comes from new knowledge challenging past solutions. The container ship-box was not a consequence of data or even of a theory but of a revolutionary change in thinking that dramatically increased world trade and wealth. Newton’s law of gravity was replaced by Einstein’s law of relativity, “not as a logically implied or predictable next step but as a shocking revolt against the most firmly established propositions of physics,” as one analyst put it. (READ MORE: Elon Musk’s AI Bot Is Funny?)

Future growth and progress require freedom from regressive rules. Growth could survive private development of AI since that still allows choice and competition from alternatives and probably some means to avoid or contain its effects. The private sector lacks real power to enforce the results produced by AI. But national government has the power and has proved easily tempted to force the results of its backward-looking solutions upon its dependent populations. So, my modest proposal would be a law to forbid the national government from using AI as a determinate for its own policies. 

Indeed, the U.S. Governmental Accountability Office has just released a report finding that 20 national government agencies are proposing 1,200 artificial intelligence use-cases, asking Congress to approve $1.8 billion to fund them. And those agency processes would be overseen by multiple government-wide management offices to spread AI to the rest of government.

The GAO report is a wake-up call for the need to place at least a temporary congressional hold on adopting these and other government AI agency plans. Or it might even take another 10,000 years to free society from this backward-looking AI bureaucracy trap.

The post Confronting the ‘Artificial Intelligence’ Backward-Looking Bureaucracy Trap appeared first on The American Spectator | USA News and Politics.