I started a thread here some time ago in an attempt to find perspective. I never did. Three years later, I finally found some in a November 19, 2018 post in Sabine Hossenfelder's Backreaction blog.
The present phase of stagnation in the foundations of physics is not normal.
Nothing is moving in the foundations of physics. One experiment after the other is returning null results: No new particles, no new dimensions, no new symmetries. Sure, there are some anomalies in the data here and there, and maybe one of them will turn out to be real news. But experimentalists are just poking in the dark. They have no clue where new physics may be to find. And their colleagues in theory development are of no help.
The self-reflection in the community is zero, zilch, nada, nichts, null. They just keep doing what they’ve been doing for 40 years, blathering about naturalness and multiverses and shifting their “predictions,” once again, to the next larger particle collider.
That looks so god awful familiar. Theories (like Global Workspace, IIT, etc.) based on wishful thinking that still can't be reconciled with evidence. Plugging away with the same limitations and methodologies that we used 60 years ago. The continuing desperate hope that the next version of Watson or Alpha Zero will at last shed some light on AGI.
The problem is also not that we lack data. We have data in abundance. But all the data are well explained by the existing theories – the standard model of particle physics and the cosmological concordance model. Still, we know that’s not it. The current theories are incomplete.
Yes. There is excellent work being done everyday in computer science, behavioral science, cognitive science and neurology. Everything we do within computer science has already been well explained with computational theory. But we know for a fact that it is incomplete.
I am merely summing up predictions that have been made for physics beyond the standard model which the Large Hadron Collider (LHC) was supposed to find: All the extra dimensions in their multiple shapes and configurations, all the pretty symmetry groups, all the new particles with the fancy names.
They were all wrong. Even if the LHC finds something new in the data that is yet to come, we already know that the theorists’ guesses did not work out. Not. A. Single. One. How much more evidence do they need that their methods are not working?
We could be talking about perceptrons, neural networks, deep learning. Money spent on "self-driving" cars that still don't work. Every single prediction that human reasoning could be nailed down with computational theory has failed miserably over and over and over. Every attempt to create brain models or reasoning models using computational theory has failed.
The perceptron goes back to 1957 and The General Problem Solver to 1959. The Lighthill Report in 1973 stated, "In no part of the field have the discoveries made so far produced the major impact that was then promised".
XCON dates to 1980. The Japanese 5th Generation Computer Project dates to 1981 and DARPA's Strategic Computing Initiative to 1983. Cyc dates to 1984 and Deep Learning to 1986. A decade later it was realized that XCON was too brittle and Cyc still couldn't reason. Even by 2001 the goals had not been met.
By 2007, DARPA was at it again with its Grand Challenge Program and a decade after that, no significant progress.
Deep systems, convolutional systems, feed forward, back propagation, and recurrent neural networks all seemed to come together in 2012. This was it. Now we are at Alpha Zero and still no path forward.
If you look at the sociology of science, bad incentives create substantial inefficiencies. If you look at the psychology of science, no one likes change.
Developing new methodologies is harder than inventing new particles in the dozens, which is why they don’t like to hear my conclusions. Any change will reduce the paper output, and they don’t want this. It’s not institutional pressure that creates this resistance, it’s that scientists themselves don’t want to move their butts.
It's much easier to work on a new algorithm than to create a new field of science. It's far easier to come up with a new spin on existing neural networks than delve into new theory.
I am afraid there is nothing that can stop them. They review each other’s papers. They review each other’s grant proposals. And they constantly tell each other that what they are doing is good science. Why should they stop? For them, all is going well. They hold conferences, they publish papers, they discuss their great new ideas. From the inside, it looks like business as usual, just that nothing comes out of it.
I agree with this too. If it were left up to Deep Mind, IBM, Microsoft, Tesla, Fujitsu, military, graduate and post-doc research we would never move forward. I used to think that this board actually gave a damn about science, that people here cared about evidence and critical thinking. Instead, the mantra of skepticism often seems to be just another tool in gamesmanship, a way to push your own biases and hopes without having to explain them.
How many times have I seen people here parrot, "Extraordinary claims require extraordinary evidence"? I suppose they feel good when they do this, but they don't seem to realize that moon hoaxers, flat earthers, and anti-vaxers use the same arguments.
If that had been the attitude then Goddard would have never bothered with laboratory experiments to prove the Ideal Rocket Equation; he would have instead insisted that Tsiolkovsky prove it conclusively before he would even discuss it. Kepler would never have worked on a sun-centered model and no one at CERN would ever have been looking for a Higgs particle. The "prove it first" chorus isn't science and never has been; proving it is science.
For me, for someone who has spent the past five years dragging the stone up the hill, perspective is a good thing to have. The lack of progress, the unwillingness to move off existing theories even when they don't work, and the inability to conceive of something new isn't just my imagination. I've got about a year to finish the theory if I want to publish in 2021. I don't even know if this is possible; maybe it isn't. But, so far, I've made more progress than anyone else in the field and I'm not stopping.
