Boy! What a Mare's Nest!
There is no way to unpack this thing in a chat-room posting. I'll give a few thoughts, but PLEASE don't take them to the bank --- maybe they'll serve as thought-stimulus and therefore add something to the Universe rather than (as this article mainly does) subtract from it.
So WITH HUMILITY .................
1). the Mediaeval lead-off comment is over-the-top (making oneself "an enemy" is precisely NOT the open-minded non-emotional psychology of the scientist, let alone this guy's cartoon concept of a Baconian scientist) yet, if you were already a student of the philosophy of science and gave the guy poetic license, you could say that the exaggeration makes a point (i.e. question everything --- Descartes would have said "doubt" --- including your own biases at each step of the investigation.)
2). author says that there is today too much data for the scientist to synthesize --- sure, a no brainer, but this has been true for quite some time. Author ascribes this to limitations of the human mind --- well, OK, but in emphasizing this the author demonstrates that he doesn't really comprehend the totality of the "method." The scientific method is never in the end about the individual human mind, but about the collective of human minds researching each subject. Whereas all individuals are supposed to behave properly (objectively, data-oriented --- "save the data not the theory") towards their work, because of their limitations (both quantitatively, emotionally, and due to the accidents of how their life proceeds to present only bits of the relevant data to them), individuals will not get it perfectly right. The Method attempts to moderate this incompleteness and bias and error by dumping all the data/conclusions into an intellectual stew-pot wherein it is scrutinized, retested, re-presented with polish or severe alterations --- think of it as a non-individualized distillation process, where many flawed individual contributions go in, and few survive intact, but those that do are (at least in part) cleansed of individual error. This public or, better, community element requires things to be published with full disclosure of how the data was achieved and how it was analyzed. Anything which does not meet this requirement is not "science."
3). This guy's inclusion of comments about the increasing errors in medical papers is an unhelpful intrusion having nothing to do with the point he tries to make, and shows his own idiosyncratic bias (he features such topics in his writings) that he would include this distracting irrelevancy.
4). He cites "chance" in a scientist's training and life (presenting that scientist with certain data and thoughts, but not all data and thoughts) as creating a deplorable individuality in science. This viewpoint is almost too much --- laughable in the extreme. It is precisely THROUGH that individuality that unique approaches to issues are created, and then it is through the aforementioned (and ignored by the author) "communal distillation" process that idiosyncratic theories and approaches are evaluated.
5). So then he launches into his solution ---> the very old concept of Baconian Robotic "scientific" discovery. We've analyzed that idea for 400 years, and, so far, it's nearly sterile. BUT we're being "scientific" here, so we give this guy a hearing. What he has to say contains no new insight at all. The theory has the non-scientific "romantic" intuition that somehow robotic collection of data will not only discover patterns in masses of data, but will somehow leap beyond those patterns to the realm of Causes --- i.e, find not only "Laws" of behavior, but defendable Reasons WHY that behavior occurred. These two "products" of the science (actually "intellectual discovery/scholarship" is more inclusive and better) are on quite different grounds when it comes to mental agility, and one would assume artificial "mental agility" as well --- you can intuit whether the AI world can even do "causes" at a sophisticated level at all.
Just for fun, not "proving" anything, but .... let's free up a few Baconian Robots to blunder into the world (remember they MUST "blunder" randomly for every "direction" given contains biases programmed by us), and we wait and see what they come up with. My guess would be nothing over centuries of time, but let's say they're really good (and lucky.) The robots come back. A few of them have fixated on "Red". They share their separate fixations. One notes that The Sun is sort of Red-Orange. One notes that cherries are Red. One notes that many apples are Red. Another, Tomatoes. Electrons spinning furiously they hit upon the idea that the Sun must transmit Red somehow to these things because they don't happen when the Sun is not strongly shining. And this is true every season. It is a Law of Nature, therefore, that The Sun can transmit Red and some but not all things receive it. Another robot then contributes the observations that if humans stay out in the Sun when it is strong, THEY turn red. Most do anyway. Once more Red is transferable. Then another robot remarks that it has been studying Pies. If you make Pies out of Cherries and Apples, even Tomatoes, and human individuals eat a whole lot of them, those humans begin looking tired and sweaty and, well, red. Not only that, but, the robot notes, The Sun is Round. Cherries and Tomatoes are Round. Even Apples are sort of Round. Pies are Round. And ... here's the clincher ... if humans eat a lot of Pies, THEY become Rounder themselves. The Baconian Robots have discovered a very "fruitful" (ouch) line of research into the Laws and Causes of Nature.
What's the point? Random directionless blundering about is not only NOT how Science progresses, but would produce a nearly uncountable number of false leads, taking centuries to track down and discard --- and, in a "pure" objective robot universe, what could be a purely objective criterion for discarding anything?
We make progress using both "halves" of our brains --- i.e. both styles of thinking: analytical AND synthetic, logic AND intuition. AI MIGHT be good (with our pre-programmed meddling) at collection, pattern-search, and looking for repetition, but intuitive "jumping" to a hypothesis to direct new angles of research?
6). With all of the above and several library shelves more, it is my opinion that the author's "three different planes" of mental activity misses reality by a wide margin. Then he adds to his restricted view of things (exactly what he's claiming to avoid) by saying that the data-gathering is by far the biggest hurdle --- wow. Precisely wrong and out-of-touch with everything real in a scientist's life as a scientist.
7). one last, as usual, incomplete remark: if you are around real scientists it becomes quickly clear that the group is composed of two wildly different types. The VAST majority of these people are "brick-makers." They take other persons' big thoughts/successful hypotheses and do some simple add-on to the edges of their work. Often they continue to make bricks all their career, never having a real creative contribution. The other rarer types are, of course, the Architects. These guys imagine, test, and rough out the Big Ideas which drive the fields. The author of the current article somehow thinks that the Architects will be readily replaced by AI brick-makers. In my view, IF AI really did take-over the science literature, it would be the Dark Ages rather than The Renaissance.
... again, very large topic, merely skimmed here. Take my views with Sodium Chloride.
