Artificial intelligence (AI) has long been held as the end for traditional jobs. Learning about OpenAI’s new language, GPT-3, raises lots of questions. Before I was optimistic about my chances of becoming a doctor. Now I’m asking a more serious question; “is my medical degree worthless?”
AI’s already advanced enough to write poetry, compose music and code web-apps with the slightest of prompts, what’s not to say it won’t stop there? Overhauling healthcare could well be next in machine learning’s sights.
GPT-3: An Introduction
Right now you’re probably thinking, besides from being a weird acronym, what exactly is GPT-3?
GPT-3 is the product of San Francisco-based research laboratory OpenAI , a non-profit turned commercial enterprise founded by Elon Musk and Sam Altman. It stands for Generative Pre-trained Transformer. The third of its kind, obviously.
Put in as basic terms as I can (understanding it is a crazy challenge itself); it’s a predictive text language model. Which means you tell it to do stuff and it’ll spit out what you ask (like write a press release for Kanye West as President, or give you a noir Harry Potter). Fun.
Take a look at some of the mad examples of GPT-3 around the web and you’ll see some interesting (and frightening) use-cases. It can generate long-form text, commentate on sports and deliver instant answers to complex multiple-choice and open-ended questions. Even the best medical technology in hospitals can’t do that.
Why GPT-3 (& AI) Is a Big Deal
I’ve been playing with the platform myself thanks to this useful backdoor-workaround. Meaning I haven’t had to apply for full API access (which is what these software engineers do) from OpenAI themselves.
What makes it a big deal is its size. GPT-3 is the world’s largest language model ever created — built on 175 billion parameters. This eclipses its previous iteration, GPT-2. And also the one built by Google (BERT) last year.
To guys like me, without a computer science brain to really understand this stuff, that’s really powerful. But not powerful enough to the really smart people at least. Those who argue we shouldn’t worry too much yet. Given it’s not capable of understanding or meaning – or out to kill us like The Terminator.
Is My Medical Degree Worthless? Maybe, but Wait
Before getting all depressed about why something like this might put me out of a job, let me first explain. My fear? Not very productive.
Even if GPT-3 was able to replace many areas of medicine; I’d still argue it’s a net good. Enabling systems to communicate with more speed and accuracy than your average clinician, it might help reduce human error. It could also diagnostically problem-solve at the drop of a hat too.
Not only would that enable medics to focus energies on direct applications of their skill-set – taking all the bureaucracy, patient reports and letters of referral out of the equation – it would also (perhaps) lead to better patient outcomes.
How GPT-3 Could Disrupt Medicine
Case in point; direct communication with a chat-bot. You go to it with your symptoms, it tells us you the next steps.
Surely that would work better, at medical triage, than a tired, over-worked receptionist deep into the 12th hour of a night-shift?
As would all the remote GP-consulting apps springing up. Suddenly staffed by competent AI doctors who could advise on useful treatment plans while placating the anxieties and concerns of the worried patient. Even deal with emergency late night calls.
Speaking to Hawaii-based Rainer Domingo, a software engineer specialising in artificial general intelligence (and the 11-year head developer at a 50 physician multi-speciality clinic), gave me even more food for thought. The future of tech in medicine could take us out of the clinic entirely.
Here’s what Rainer had to say.
In a clinical setting, I envision [there being] a digital physician’s assistant that can do comprehensive patient intake, diagnosis and patient QA even after the patient has left the office. The digital PA can also take care of follow-up visits after the patent has seen a physician. I think it’s well within the realm of possibility that a physician may not have to actually see a patient unless a patient requires physical contact or visual inspection of something the computer vision system cannot see.
If true, that’s some pretty amazing stuff. But where does that leave us?
The Human Touch: A Doctors Saving Grace?
The other side of the coin is also worth looking at. What happens if artificial intelligence, like GPT-3, is more effective than Ranier suggests? Forget not being in clinic, would the role of a physician even exist at all?
Obviously there’s a strong case, at least in most medical specialisms, for that not to happen. Reddit posts like these (what makes a good doctor) — that point to communication skills and ‘being nice’ as the key assets — make an important point. Patient-doctor interaction is still one of the major faces of the industry. Still one necessary to get right. And not one that can be done very well by a robot.
Where else AI could stumble is with the more mechanical aspects of medicine. GPT-3, right now, can’t pair with a robotic super-arm fixing broken bones on the surgeon’s table. Nor does it prove useful, as far as I’ve seen, running its eye over several thousand-degree magnifications of biopsied tissue in the path lab.
Right now GPT-3 also makes, as OpenAI CEO Sam Altman says; “silly mistakes” in its attempt to deliver what its puppet master demands of it. It’s also plagued by issues of bias. Even reported – hardly surprising given all the web forums and Reddit it’s scraped – to be a little racist and sexist too.
But then so are some doctors. Aren’t they?
In conclusion then it’s too early to tell what’s going to happen with languages like GPT-3 and particular fields of industry. One trend we can expect however is AI’s continued rate of expansion and refinement.
In terms of education-based decisions this opens up some good talking points. Especially those made by people going into (or already in) medicine keeping one eye on the future. The real question they want to be asking, as I see it, is this:
“Would I still want to go into a five or six-year specialist training program now for something that’s likely to be overhauled by tech in the next couple of years?”
Given the shortage of physicians, increasing health care costs and an increasingly older population, there will need to be more drastic solutions. And AI will be at the forefront of that troubleshooting.
One exciting caveat for when it does though? Being able to use the technology to eliminate the mundane or repetitive aspects of our jobs . And give us an opportunity to expand further on our roles as simple “clinicians”. Possibly taking us deeper into the tech and software worlds than we would have otherwise expected.
As for what to do in the meantime while we all sit around and wait for AI to hit us?
Use this article as a reminder to seek to understand more about the world of machine learning. Tell yourself your medical degree isn’t worthless. Get working on improving the human side of things.
When the robots really do come and steal our jobs, you’ll be needed.
Interested in finding out more? Check out some of the best GPT-3 use-cases here.