The Singularity Is Near

One might say The Singularity Is Near:

AI Pores Over Old Scientific Papers, Makes Discoveries Overlooked By Humans

Researchers from Lawrence Berkeley National Laboratory trained an AI called Word2Vec on scientific papers to see if there was any “latent knowledge” that humans weren’t able to grock on first pass.

The study, published in Nature on July 3, reveals that the algorithm found predictions for potential thermoelectric materials which can convert heat into energy for various heating and cooling applications.

“It can read any paper on material science, so can make connections that no scientists could,” said researcher Anubhav Jain. “Sometimes it does what a researcher would do; other times it makes these cross-discipline associations.

The algorithm was designed to assess the language in 3.3 million abstracts from material sciences, and was able to build a vocabulary of around half-a-million words. Word2Vec used machine learning to analyze relationships between words.

“The way that this Word2vec algorithm works is that you train a neural network model to remove each word and predict what the words next to it will be,” said Jain, adding that “by training a neural network on a word, you get representations of words that can actually confer knowledge.

As one example, researchers fed publications from before 2009 into the algorithm and were able to predict one of the most effective modern-day thermoelectric materials four years before it was actually discovered in 2012.

The technology isn’t restricted to materials science either – as it can be trained on a wide variety of disciplines by retraining it on literature from whichever subject for which one wants to provide a deeper analysis.

“This algorithm is unsupervised and it builds its own connections,” said the study’s lead author, Vahe Tshitoyan, adding “You could use this for things like medical research or drug discovery. The information is out there. We just haven’t made these connections yet because you can’t read every article.”

One could also say, with a similar amount of justification, Skynet smiles.

7 thoughts on “The Singularity Is Near

  1. I, for one, welcome the arrival of our machine overlords. They’ll likely rule with more predictability than current liberals.

  2. I presume that the punchline will be that someone broke it and rendered it forever useless by feeding it 1990s postmodern literary criticism articles and doctoral theses in 20th Century English Lit.

  3. I’m not concerned. Even if AI can sift over old papers and find something new, I’m not convinced that there is that much knowledge still to discover.

    In the 1990s we were going to unlock the secrets to life via Genetics and DNA analysis. By the 2000s, we realized that path was not going to unlock any secrets except in special applications. It was a great disappointment to the researchers.

    Likewise, efforts to cure cancer, heart disease, metabolic disease, kidney disease, social behaviors, and so on have stalled with diminishing returns.

    I suspect that there is a Heisenberg’s uncertainty principle that applies to the limits of science. AI cannot discover what is undiscoverable. What’s left to discover in Mathematics, Physics, Statistics, Biology, or Pyscology? It’s true, AI can sift through more papers, but most breakthroughs rapidly spread via word of mouth anyway.

    For example, there are 40,000+ papers per year published on diabetes, but those are mostly noise from the academic community. And we still don’t understand diabetes. About all that we agree on is that is defined by high blood sugar. There is no cure! Perhaps there really is not a cure.

    Think about this. My wife just when through a stage 3 ovarian cancer treatment that looks like it will be successful, but it used the same techniques that have been available for 30+ years: surgery and chemo. It’s been six months since her last chemo and she seems OK. But in reality, even the Doctor does not really know if she still has cancer. Tumor markers are not reliable and imaging will not help. How will AI help? Sure it’s possible that a ‘cure’ will be found, but it is also possible that the tooth fairy is real.

    So yes AI may help, but I expect to help only at the margins.

    • Disagree. I think great breakthroughs have been made, such is in genetics and disease. Just not released and publicized, because they’d shut down entire treatment-related profit streams; can’t sell a decade’s worth of daily chronic pharma if you can treat it with single shot. Also, some were very non-PC, and shot holes in cherished left-wing dogma, so they were circular-filed because they’d cause feel-badz; god forbid we actually allow effective teaching in schools, and stop wasting resources on ineffective programs for people unwilling / unable to learn.

  4. Naaa…

    It’s just a program looking for associative properties in a vast sea of data.

    I’m sure that they didn’t publish all of the misses and mistakes it made.

    Machine learning is still GIGO.

    A recent article about this “super smart” AI discovered that it wasn’t analysing the photos it was given by the actual content but by the watermark in them. It “cheated” it’s way to the results that the operators wanted.

Comments are closed.