Is AI really going to kill us all? Well, that’s the buzz since a junior staffer quit Anthropic and announced the end of the world. And he may be right. But, not today..Tonight’s guest on Hannaford is Western Standard reporter Leah Mushet, who has actually been living in the story and who a few days ago came up with an astonishing account of what people are doing with AI now – not the science-fiction horror version, just the version already in hospitals, newsrooms, and even kitchen counters.Most of us have poked at Grok or ChatGPT. We know it can draft a paragraph, search local breakers’ yards to find a spare part for an old car, or invent a recipe out of whatever is left in the fridge..Leah has been following something harder. Researchers have taken the kind of systems that sit behind those chatbots and put them into da Vinci robots to perform advanced surgery. The robots watched gallbladder procedures, learned from the footage, and then guided the surgery themselves on pig cadavers. The published claim is a hundred percent success rate on those trials. No surgeon’s hand on the instruments. Arms, tools, computer, model.Read it here..AI SAVING HUMANITY? Six ways AI is helping the world right now.That raises the obvious question: if they can do it on a dead pig, what stops them doing it on a living person? Leah’s reporting is that we are not there yet. Maybe soon though, under the supervision of a surgeon; gallbladder work is common..Brain work however, is not. At University College London however, it has started in another form. A team used AI on live footage during a pituitary tumour extraction. The pituitary is a small, hard-to-get-at gland at the base of the brain. Nevertheless, a programmed AI guide advised the surgeon where to cut, much reducing the risk of vision loss to the patient. That is not a robot operating alone. That is a machine watching the field and talking to the human holding the knife. The line has already been crossed. The only argument left is how far and how fast.Ah you say, what about the young man who left Anthropic after four months and told the internet the technology was going to kill us all. Jacob Coxon’s post went everywhere. Other people in the field nodded along in the vaguest possible language – powerful now, dangerous later..Press them for the mechanism however, and you get generalities. Replacement. Outsmarting us. Systems that lie to users. That lawyer who filed fake cases invented by a chatbot… He has become an urban legend. (Though not a myth. This happened.)For our comfort though, that is not a robot pulling a lever on a dam or interfering with communications and causing airliners to crash. It is just a tool that sounds sure when it is guessing. Which is all it can do. And with enough training, it can do that well enough. But there are things it will never do. As Gleb Lisikh argued in a C2C article a few months ago, you “can’t make pigs fly by strengthening their leg muscles.”Leah uses these systems the way a lot of reporters now do: to hunt data, ask for the links, then go back and check, check, check. Ask for the links. Go back to the hospital release, the paper, the original document.“If your mother says she loves you, check it out.”That rule did not expire when the software got fluent.So tonight we are going to walk the useful parts and the scare parts without treating either as a sermon. Surgical robots that copy good surgeons are useful. AI that whispers during brain surgery, works. Doomsayers who will not specify the disaster. Search engines that answer the question and starve the website that did the work.There is just this. As I watched a recent grad who had been allowed to use calculators at school struggle to calculate 5% GST on a $100 purchase, I did wonder: if AI becomes our first stop when we want to know something, how long before we cannot do the simple things ourselves?The Hannaford Show uploads at seven o’clock.
