Artificial Intelligence & Its Future Role in Science
Throughout the summer, I have read columns by Joel Achenbach, a columnist for the Washington Post who mainly discusses a wide range of science topics. Within his writing, he attempts to present unbiased facts on the topic at hand. He often presents an optimistic view of the newfound knowledge he delivers to his readers. This can be seen in his article: “See the sharpest image yet of a supermassive black hole.”. This column beautifully details his utilization of tone and the establishment of his personal writing style.
Within this article, Achenbach describes a newfound discovery pieced together by astronomers this year. This photo was created by combining recent technologies, with the main image originating from the Event Horizon Telescope, “a consortium of telescopes across the planet”. However, the completed image could not have been generated without the help of PRIMO, a machine-learning algorithm, or, in simpler terms, an AI. This artificial intelligence assisted in filling in the gaps that the telescopes were unable to pick up. Utilizing a well-made, but limited, data set, this machinery helped expand our understanding of how black holes may look. Here is where the issue arises: PRIMO relies on computer models to complete this image. This means that this image is almost entirely reliant on our understanding of what black holes look like. Thus, it is difficult to say whether it is entirely accurate, with it requiring more verification when we eventually receive more data. This, nevertheless, could take many, many years.
Discussion Question: In situations such as these where data is limited, should artificial intelligence be trusted to expand upon the knowledge we already hold? Do you believe that AI will be helpful in creating new scientific discoveries in the future, or should this job be left to us?
I think that in these situations that A.I. is definitely a useful tool to get the results that we want like how they used it to make the photo of the Black Hole. However I do feel that we should use some caution in its use. When training these A.I’s we should be careful to only put in info that we know for sure instead of our own bias. The proliferation of A.I. in our society today will certainly lead to many new scientific advances but both the scientists who use the programs and the reporters who talk about it need to remember the role that the A.I. played in the discovery and stress that to the public and make sure that the A.I. can’t have taken much influence from the humans who made them.
While AI definitely has some practical uses that could greatly benefit the progression of scientific discoveries, I believe that it should be used with an abundance of caution. For instance, AI has the ability to process and analyze vast amounts of data quickly, which can accelerate scientific research. It also has many practical applications, such as handling repetitive tasks and large datasets. In this sense, AI could greatly benefit scienfitic discoveries. However, I don’t think that AI in and of itself could be used to make scientific discoveries. AI’s intelligence is entirely based upon the data that it has been given, meaning that it is, in actuality, not creating anything, but rather repurposing everything it has been fed to make it appear like it is creating something. While this in and of itself is not harmful, limited data sets are more likely to be incorrect, meaning that an AI that is fed a limited data set is more likely to be incorrect, and the danger of presenting falsehoods to be discoveries could be detrimental.
While A.I. can be beneficial for calculations and organizing and analyzing data, having it do most of the work is currently very risky and the technology we have is not sophisticated enough to be fully trusted. So we should use technology and A.I. in some ways, and we do almost everyday day, (For example, using a calculator to do our calculations for us) but it shouldn’t be trusted for everything. At the current state, we should only use A.I. to confirm or check what we think we know, and eventually, we might be able to use A.I. for just about everything in the world of science. Overall, we need to exercise caution when using A.I. as our technology is not yet advanced enough for that, but it can still be useful in many ways.
In situations where our data is more limited I believe that A.I. should be viewed as an educated guess at most. I think that we put too much faith into the power of artificial intelligence. As stated before A.I. just uses previous examples and compiles them together. This means that it isn’t really creating much of anything new and coming to its own conclusion. It’s just assuming things based on what we think is right at the moment but could be proven wrong eventually, like most things are. I believe that A.I. could be used to look at possible answers but I don’t think we should be so gullible as to assume their statements are true when they are also just assumptions.
When data is limited, I believe that artificial intelligence should be used to help, but not to expand upon, current knowledge. It is not advanced enough to be something we can rely on yet. In the far future, it could have potential to lead to discoveries we never would have made on our own. At its current state, it is not something we can trust completely. It should be used to help people, but not to expand on what they already know. It still has potential to make mistakes. AI only has access to the knowledge that we have. It doesn’t know anything that people don’t already know, and it can’t do anything that people can’t already do. We should use it for helping with things and checking mistakes, but we should not use it for something as important as expanding current knowledge on topics. And if we do use it for that, we should understand that what it is telling us may not be entirely true. However, it should also be noted that AI has potential to point out things that we may not have noticed.
When it comes to these circumstances, I believe that AI should be trusted but only when needed. Using an AI can be very useful as work can be done efficiently without much effort. Especially if we don’t have enough knowledge or time to do a task, it can come in handy. However, it’s not too reliable as AI takes in information that’s already been said or mentioned which could create false accusations. Artificial intelligence should just be used when it is appropriate like how your article mentioned making an image to show what black holes may look like or digitally organizing a data table in seconds. With the hand of AI, it can certainly be helpful to making scientific discoveries in the future but the job should still be left for us to do as we cannot always rely on this tool.
AI can definitely be trusted only when it comes to broadening the knowledge we already have. AI can definitely do a lot of things better than us and shorten the time things can be done by but it doesn’t mean it can be trusted with gathering information even we don’t know about. Some of it could be completely false and end up making use panic for no reason. AI can eventually be useful when it comes to making new discoveries, but for now it should just be us making the discoveries until we truly hone the potential of AI when it comes to information gathering.