I have used ChatGPT to double-check complex diagnoses conducted on myself by real doctors, and I am overall satisfied by its behavior. In limited cases, ChatGPT went beyond and suggested me new viewpoints, new paths. Of course, I was constantly en-garde, never trusting the tool and asking trick questions, citations, verifications.
I believe that neither an experience like mine nor one like doctor JT-S 's is definitive for judging the quality of LLMs as physician's aids. We need more unbiased experiments and we need more field experience.
Before we test, we also need to fine-tune the LLM on medical knowledge, which I don't think is the case with ChatGPT.
We also need to benchmark AI tools against human physicians (acting as consultants) rather than an unexistent ideal...
WSJ CIO Journal had an interesting opening story. I'm quoting their newsletter summary. For the full goods, read https://cio.cmail19.com/t/d-l-vmtyz-irtduikulr-k/. Then come back and read this piece. SoftBank Group CEO Masayoshi Son's use of ChatGPT mirrors what this note talking about.
"Good morning, CIOs. Many executives remain excited about the potential of ChatGPT, but wary about fully integrating the technology, citing concerns over its use of online data, security risks and its grip on reality.
And then there is SoftBank Group CEO Masayoshi Son. At the annual meeting of the Japanese technology-investment company on Wednesday, Son told shareholders he has used ChatGPT every day for brainstorming, The Wall Street Journal's Megumi Fujikawa reports.
Those sessions, some into the wee hours, resulted in more than 600 ideas. He described one exchange—at around 3 to 4 in the morning—in which he pitched an idea and then answered the AI chatbot as it raised objections. “After we repeated this several dozen times, I really felt great because my idea was praised as feasible and wonderful,” he said.
Reality. Losses from its technology funds had left SoftBank with a net loss of $6.9 billion for the fiscal year ended March, the company reported in May. It was the company’s second consecutive year in the red.
ChatGPT, CEO whisperer. “The time is approaching for us to go on the counteroffensive,” Son said Wednesday. “I want SoftBank to lead the AI revolution.”
And Andrew Ross Sorkin's daily column in The New York Times has a similar story about Son.
"But as the world races to embrace artificial intelligence — something that has long fascinated him — Son used his company’s annual shareholder meeting today to publicly, and memorably, restate his commitment to becoming a leader in the cutting-edge technology.
“We are ready to shift to offense mode,” Son told investors and analysts, adding: “I want SoftBank to lead the A.I. revolution.” Reviving the kind of grand presentations that he long favored — who can forget the “valley of coronavirus” slide in 2020, featuring flying unicorns? — the SoftBank chief began by asking, “What is Mankind?”
The answer, apparently, is something that can benefit from the technology behind chatbots, which has already spurred an investment boom. (To be fair, the potential of A.I. was one of several central theses behind SoftBank’s $100 billion first Vision Fund.) “When your grandkids are our age, I believe they will be living in a reality where the computer is 10,000 times smarter than the sum of all human wisdom,” he said today. Where SoftBank would fit in, he added, was to try and “become an architect to build the future of humankind.”
Son argues that SoftBank has better days ahead. While the company’s Vision Funds have suffered big paper losses amid the decimation of start-up valuations, he said that SoftBank has since amassed over $35 billion in cash in “defense mode,” and was ready to invest it. (The company also stands to benefit from the forthcoming I.P.O. of Arm, the chip designer it owns.)
And Son said he’s excited again, after lying low last year. “There were times when I felt so empty,” he told investors. ‘Is this enough? Is this it?’ I cried and cried and couldn’t stop crying for days.” Now, he added, “I’m having too much fun.”
I have used ChatGPT to double-check complex diagnoses conducted on myself by real doctors, and I am overall satisfied by its behavior. In limited cases, ChatGPT went beyond and suggested me new viewpoints, new paths. Of course, I was constantly en-garde, never trusting the tool and asking trick questions, citations, verifications.
I believe that neither an experience like mine nor one like doctor JT-S 's is definitive for judging the quality of LLMs as physician's aids. We need more unbiased experiments and we need more field experience.
Before we test, we also need to fine-tune the LLM on medical knowledge, which I don't think is the case with ChatGPT.
We also need to benchmark AI tools against human physicians (acting as consultants) rather than an unexistent ideal...
WSJ CIO Journal had an interesting opening story. I'm quoting their newsletter summary. For the full goods, read https://cio.cmail19.com/t/d-l-vmtyz-irtduikulr-k/. Then come back and read this piece. SoftBank Group CEO Masayoshi Son's use of ChatGPT mirrors what this note talking about.
"Good morning, CIOs. Many executives remain excited about the potential of ChatGPT, but wary about fully integrating the technology, citing concerns over its use of online data, security risks and its grip on reality.
And then there is SoftBank Group CEO Masayoshi Son. At the annual meeting of the Japanese technology-investment company on Wednesday, Son told shareholders he has used ChatGPT every day for brainstorming, The Wall Street Journal's Megumi Fujikawa reports.
Those sessions, some into the wee hours, resulted in more than 600 ideas. He described one exchange—at around 3 to 4 in the morning—in which he pitched an idea and then answered the AI chatbot as it raised objections. “After we repeated this several dozen times, I really felt great because my idea was praised as feasible and wonderful,” he said.
Reality. Losses from its technology funds had left SoftBank with a net loss of $6.9 billion for the fiscal year ended March, the company reported in May. It was the company’s second consecutive year in the red.
ChatGPT, CEO whisperer. “The time is approaching for us to go on the counteroffensive,” Son said Wednesday. “I want SoftBank to lead the AI revolution.”
Read the story at https://cio.cmail19.com/t/d-l-vmtyz-irtduikulr-k/
And Andrew Ross Sorkin's daily column in The New York Times has a similar story about Son.
"But as the world races to embrace artificial intelligence — something that has long fascinated him — Son used his company’s annual shareholder meeting today to publicly, and memorably, restate his commitment to becoming a leader in the cutting-edge technology.
“We are ready to shift to offense mode,” Son told investors and analysts, adding: “I want SoftBank to lead the A.I. revolution.” Reviving the kind of grand presentations that he long favored — who can forget the “valley of coronavirus” slide in 2020, featuring flying unicorns? — the SoftBank chief began by asking, “What is Mankind?”
The answer, apparently, is something that can benefit from the technology behind chatbots, which has already spurred an investment boom. (To be fair, the potential of A.I. was one of several central theses behind SoftBank’s $100 billion first Vision Fund.) “When your grandkids are our age, I believe they will be living in a reality where the computer is 10,000 times smarter than the sum of all human wisdom,” he said today. Where SoftBank would fit in, he added, was to try and “become an architect to build the future of humankind.”
Son argues that SoftBank has better days ahead. While the company’s Vision Funds have suffered big paper losses amid the decimation of start-up valuations, he said that SoftBank has since amassed over $35 billion in cash in “defense mode,” and was ready to invest it. (The company also stands to benefit from the forthcoming I.P.O. of Arm, the chip designer it owns.)
And Son said he’s excited again, after lying low last year. “There were times when I felt so empty,” he told investors. ‘Is this enough? Is this it?’ I cried and cried and couldn’t stop crying for days.” Now, he added, “I’m having too much fun.”
I distrust a bit the judgment of all those who invest in AI after the publication of ChatGPT in October 2022 😉