Sure, I see plenty of uses for ChatGPT in their field, but to not even check the output before submitting it is what blows my mind.
Also I wonder how well the training data was cleaned from all potentially harmful stuff like for example the sovereign citizen ideas...
You really need not wait that far if you have been observant enough, they are they were at one point clownish enough to believe in religion, now the vast majority of them believe in 'science' etc. Over reliance on Chat GPT is just the one of the newest manifestation of laziness and dumbness.
But also, pretty sure Doctors and Lawyers are among the least trusted/highest corruption professions. All that schooling is a barrier to entry to keep outsiders away, it doesnt guarantee competence. Basically once you get in, graduation rates are close to 100%.
My wife runs the resident eduction for her division and they are very aware of the dangers of sending an incompetent doctor out into the world to practice. They’ll force them through extra training and make them repeat residency if they have to.
Residents also drop out of their programs and move into other ones. It’s rare for them to drop out completely because they’d still be on the hook for $200k+ in loans. But medicine is a big field and it’s rare that someone who made it into and through med school to not be able to find something they are competent and capable at. Maybe you’re not great with patients, but you can look through a microscope all day as a pathologist etc…
Incompetent and corrupt doctors still make it through the process of course, but far far fewer than in our profession and just about any other profession I can think of.
As for lawyers:
Law schools have around a 77% graduation rate nation wide. And about 90% of law school graduates eventually pass the bar exam.
So we’re look at somewhere around 70% of people making it through after a pretty selective filter to begin with.
n=1 anecdote here:
When I was in research track biochem and bumped into premed students, nearly the entire population of premed didn't seem to care about the science behind what they were learning. They formed a very different social clique. The "top students" in that set were sharing previous years' organic chemistry exams and none of them would read papers or get involved in research. They were entirely disinterested in the science.
In my interactions with general practitioners, I like to talk about medicines' method of actions, pharmacology, the actual biochemistry behind diseases. Most of them seem to have no clue or retention of this information. I'm not trying to challenge them, either -- I'm generally curious to learn.
That's not to say all doctors are like that. A lot of the surgeons and specialists I know are walking tomes of information.
Once you get to medical research groups that formulate those protocols, that’s where research becomes important.
Isnt this even more necessary for them to learn the fundementals? Especially if Physicians claim that they are both Art and Science, which is why we can't merely use science for diagnosis and treatment.
So they do learn fundamentals but fundamentals do not come out of research papers.
When I took non-required theoretical classes, I tended to be one of the only undergrads in the class
Also most other countries have national education programs for secondary school. High school in the US is regulated at the state and local level, so it’s very variable. As a consequence, the first 2 years of college tends towards general and education.
Another note, there’s also a med school entrance exam in the US.
Agreed regarding the US high school standards but students that take AP courses in high school can take the MCAT. They may not do the “best” but they will know some of the material. Schools in the US like UMKC have 6 year programs were high school students take the MCAT.
That’s a very different sample. Then you have to consider that the people that you are most likely to notice are the most vocal self identified “premed” students since most colleges don’t actually have premed degrees.
As far as interactions with doctors. GPs are the least likely to get into specifics because of the generalist nature of their work. Expecting them to remember something that they haven’t used in 10+ years is asking a lot.
Understanding calculus is incredibly useful to software engineers. But try asking a practicing software engineer 10 years out of school some textbook calculus questions.
https://www.opensecrets.org/federal-lobbying/top-spenders?cy...
They are the 4th largest briber in the US.
Finding a difference between bribery and federal lobbying with money, is the same outcome.
Money goes from person who wants, to person with power who receives. Everything else is theatre.
The money that comes from lobbying is also limited and pays for reelection, the candidate doesn’t get to use it for personal expenses.
There are clearly problems with the system, but yelling no! it’s exactly the same as taking bribes isn’t helpful.
Lobbying is legal because we live in a society that places such a high value on money that we've accepted bribery as a means to increase its power.
With that information I think it's less mind-blowing, they were simply assuming ChatGPT could do the work of a paralegal
Of course after some number of iterations, AI may gain the ability to self-check and self-correct to the point of achieving 100% accuracy, but I think that milestone will quickly lead to AGI and job security will quickly become an obsolete topic for every industry anyways.
It's because, just like human memory, they aren't databases or search engines. Generative transformer models are basically next-word-prediction machines on steroids. They take the input and try to "guess" the most likely reply based on their training data.
These machines have no way to distinguish facts from fiction, only probabilities of combinations of words that would make the most plausible reply.
> Is it possible to prevent it?
There are methods to prevent this by incorporating specialised knowledge databases into the training material of these models. This, however, only works with models that have been finetuned on very specific tasks and topics [1].
Other approaches use AI to transform human questions ("bag of words" inputs) to queries into structured knowledge bases, match the results (e.g. tree-like structures of context and facts) to the question and turn them back into human language [2]. The downside of these methods is that they're currently limited to simple QA formats and won't feel as "natural" as talking to chatbot and requires carefully prepared and curated knowledge databases.
[1] http://jens-lehmann.org/files/2019/iswc_bert_simple_question...
This reminds me. Whatever happened to IBM’s Watson? It’s not a LLM right? But it seem similar enough function wise.
If you mean that some lawyers use paralegals to do word searches for potentially-relevant case law (i.e., published judicial opinions as precedents), that might be true. But in my experience, good lawyers either do the raw research themselves, or at a minimum they use junior lawyers to do the initial searching for precedents; then if necessary, they do supplemental searching themselves, based on their greater experience.
Did they risk it? Even after it being obvious the lawyer was lying and trying to cover up, and not acquiescing to their wrongdoing even after all of it, all they had to pay was $5k.
They still risk being disbarred (not just hypothetically, it might actually happen) + taking a massive hit to the reputation (their own + the firm). The latter one alone is a strong punishment, given how widely-publicized this incident was in the media. It cost the firm a lot of potential clients, and the lawyer might struggle with finding a decent position after the fact (given how much more reputation-based the legal field seems to be compared to something like engineering).
Now many, if not likely most, of the well-schooled are generally intelligent, but of course there are also quite a few rich idiots as well.
(Today, of course, most of the latter go into tech.)