Apr. 7—Those of us who are computer engineering-ignorant are told that artificial intelligence will always need a human in the loop to control the accuracy of AI’s output.
But that same group of us have seen way too many movies where robots take over mankind and kill the human programmer who was to control them.
That doesn’t sound too far-fetched amid the rise of chatbots, whose job is to simulate human language. Developing from chatbots, ChatGPT can mimic human writing; it can eliminate the need for manual labor.
GPT-4 has shown it can perform like a human and solve difficult problems; it passed a simulated bar exam with a score around the top 10% of test takers.
It shouldn’t be too many generations away for chatbots to take the roles of not only programmers but other workers with “tedious” jobs.
Researchers at McKinsey Global Institute predict that more than 45 million U.S. workers could be displaced by automation by 2030 amid innovations in artificial intelligence.
Instead of surrendering, let’s pause. Or so says more than 100 AI experts who wrote a letter asking labs around the world to pause development of large-scale AI systems for six months. They are concerned — fearful — over the “profound risks to society and humanity.”
They are calling on AI labs to pause training on systems more powerful than GOT-4. If the labs can’t accomplish, then the government should issue a moratorium, the letter authors write.
Those experts include Indiana University professor Lawrence Moss whose fields are in computer science and applied logic.
It sounds logical to pause. We’re all fearful of where AI research will lead. But there is also a thrill to the research, hope for progress that benefits humans and a necessity to establish America as a world leader in AI.
Recently, Indiana University hosted an AI seminar as students learned about prompt engineering, where an AI task is embedded as a question. One of the examples: “There are so many pweople who thinks that #ChatGPT will break internet. But I am scared it’ll finish what Facebook started and will break Democracy.”
Seminar participants were to pick an emotion expressing the sentiment of the comment, such as fear, anxiety or alarm. A superbly crafted prompt can enhance how AI performs. Prompt engineering currently requires trial and error.
At Purdue University, Provost Patrick Wolfe, in the presence of new President Mung Chiang, asked the University Senate how could the “judicious” use of AI, and ChatGPT, among others, help the university become national leaders in teaching and learning.
This fall, Purdue will begin a course in which students, working with an Intel cloud, can tackle full-scale AI projects. Purdue’s Elmore Center for Uncrewed Aircraft Systems is to use AI in transforming aerospace technologies. Much of the future is being explored at Purdue.
So, no, research universities and corporations are not going to pause to appease experts who have true concerns about the direction of artificial intelligence.
Americans are faced with technological challenges from other countries. We need to foster innovations while maintaining ethical research that puts people first.
Now if some researcher can just keep spam from getting into our email inboxes …