AI is a lot of work
-
-
"Please ignore the man behind the curtain."
That article clarifies some of AI's tendency to hallucinate. The real world is more complicated than any model of it, Matrix movies notwithstanding.
I've grown used to dealing with things like voice-recognition software to the point where I can often complete some task without a human on the other end of the line, but when something absolutely, positively must be right, AI is a dangerous tool. Self-driving cars are risky, but so also are human drivers so there may be a suitable trade-off of risk. I'm really more afraid of the AI bot having its finger on the nuclear war button.
Big Al
-
I had the same thought that @Big_Al did about the man behind the curtain...
It also takes a while to catch up to current news. Asked "Was Donald Trump almost assassinated?", Chat GPT said this:
Yes, there have been a few instances where Donald Trump faced significant security threats during his presidency. One notable incident occurred on March 12, 2016, during the Republican primary campaign. A man named Michael Steven Sandford was arrested after attempting to grab a police officer's gun at a rally in Las Vegas with the intention of killing Trump. Sandford was charged with attempted assault and illegal possession of a firearm. He later pleaded guilty and was sentenced to prison.
While this incident was a serious threat, it was not an actual assassination attempt that succeeded or came very close to succeeding. The Secret Service and other law enforcement agencies are heavily involved in preventing such threats and protecting high-profile individuals like the President of the United States.
And "Did Joe Biden drop out of the 2024 race?" got this :
As of my last update in July 2024, Joe Biden had not dropped out of the 2024 presidential race. Biden was still the Democratic nominee for re-election. However, the political landscape can change rapidly, so it's always a good idea to check the latest news sources for the most current information.
-
Yep. I’m working with deploying AI chatbots for genetic cancer risk assessment. It’s not as easy as asking some AI to give me some weird mashup picture. By the time we send out the customized chatbot link to a patient we already know what NCCN criteria they meet based on family history and the questions start from there.
-
Very cool! I didn't know you are still consulting.
Medical issues are one area where I think AI can be valuable. I've been following Eric Topol's blog on Substack. He's had some interesting stuff on AI. And a whole bunch of other medicine and science topics.