by Lainelle Brazee
In my limited experience with AI, the phrase I’ve heard over and over is simply, it’s helpful. It’s productive, it’s time efficient, and in most cases, it’s free– free, of course, in terms of monetary condition, not the cost of personal data collected or stored. The people of the anonymous Padlet conducted by Dr.McCoy in class seem to agree. “I sometimes use AI to summarize really difficult and complex research papers or scholarly articles that are given to students in class. It really helps…” one user says. Another has something similar, “(Chat GPT) is a really helpful tool…it’s like a little online unpaid intern”. Here, I think, is a good spot to acknowledge that AI use can cause a world of hurt, whether it’s plagiarism, reliance, or simply giving false information. If you as the reader (and I as the author) can choose to explicitly focus on the positives just for the few moments it takes to read this paper, perhaps we can set aside the catchall phrase “AI is bad” and its entailment.
AI’s functionality and practicality have everything to do with the atmosphere in which it is used, and the appropriateness of the aforementioned. For example, you wouldn’t necessarily think that AI has a use in the general workplace- but as I read there are several AI ‘super users’ who have trained the system to their benefit. Their use of AI is far different from, as Maggie mentioned in class on Friday, high school English students who delight in copy/pasting their essay prompts and the subsequent return into ChatGPT and the Google Doc containing ‘their’ essay. Kanika Khurana is a so-called super user who put time and effort into training her AI. “The 30-year-old former head of design at ProCreator, a design agency in India, uses it for research, to help write up meeting minutes from AI-generated transcripts, and to summarize her dictated thoughts after job-candidate interviews, then format them for human resources. Her team has also created custom bots that can help other (especially more junior) colleagues create content for clients using specific lingo and voice, catering to target audiences”, Danielle Abril notes in her article. If this same mindset were to be adopted across other disciplines, I could easily see how the indoctrination of AI in the workplace could be both constructive and common.
There are a lot of (mostly) baseless claims living on the internet about AI and the positive benefits it’s having on our society. If you ignore a particular half of the pros/cons chart, it’s easy to get really excited! Firstly, we’ve talked in class about the use of AI in the medical world. If something like VisualDx could have a sibling specifically dedicated to cancer recognition, for example, accurately and consistently, that would be a huge deal in the medical community as well as for patients globally. Of course, we still might have to keep rulers hidden out of sight in order to ensure accuracy. Then, there’s the problem that our AI systems are implicitly biased- wonder how that happened. There are solutions, as Donna M. Christensen, Jim Manley, and Jason Resendez point out in their article. Simple measures such as a “domain-forward approach” which involves an incorporation of doctors and health care professionals into the algorithm are all it would take to create a more fair and accurate system. Examining and lessening implicit bias in the people who drain the AI model (duh) and recognition by the Federal Drug Administration and the Federal Trade Commission of the fact that AI algorithms and machines could have potential harms as well as clear benefits. I feel as though we could be on the verge of something amazing- but there are some fatal flaws that have to be fixed before the good can completely overshadow the bad. However, the nagging voice in the back of my head has been incessantly reminding me we can’t fix problems to improve the quality of AI overall if no problems arise- the benefit of the doubt should be given. It’s simply impossible to create something that firstly immediately is foolproof and secondly stands the test of time without requiring the usual small adjustments and upgrades. Sometimes, terrible things need to happen for us as ignorant humans to realize the problem in the first place. (I will acknowledge that as I write this, I’m running full speed away from Jiminy Cricket. So if there are typos it’s because it’s hard to run full speed and hold a laptop and type simultaneously.) As the world mourns the loss of Sewell Setzer III, I’m sure some people are only now realizing just how dangerous AI could potentially be. He shouldn’t have died. But his death has not gone unnoticed, and it has brought an enormous spotlight on not only Noam Shazeer and Daniel de Freitas but on AI in general.
The fact that an inhuman model is putting white Americans over black Americans on a transplant list solely based on the fact that they’re white is asinine. Especially because something that’s supposed to be based on so many numbers and statistics that it should simply *know* that, for example (thanks Wikipedia), Hispanics alone accounted for 51.1% of the country’s total population growth between 2010-2020. The fact that a COMPUTER- a machine, something that does not have skin to color- takes that into consideration is a sad reflection of not only implicit bias in our country but also the person(s) who trained it. I’ll take the ruler cancer. However, as I mentioned earlier, the idea of AI being a successful tool in the cancer detection world is something to look forward to, and for now, we’ll allow it the benefit of the doubt.
The cord phone sexytime hysteria discussion made me internally cringe, wondering if we as a society are expecting too much of AI in its ‘infancy’ (as Mari coined). AI is a part of our future whether we like it or not, and it’s up to us to become as educated as we can about the pros and the cons. As with everything else, it needs two things to advance and better itself: trial and error, and time. Those won’t come without the latter- so all we can do in the meantime is revel in the positives and learn how to adjust the negatives. As with anything and (almost) anyone else, AI deserves the benefit of the doubt.