Major 2020 A.I. Milestones

reading time Reading Time: 6 minutes

We’re not flying in airborne vehicles like the Jetsons predicted we would be by 2020, but we do have robots and artificial intelligence – or A.I. as it’s abbreviated for convenience – has factored into robotic technology very heavily over the last few years. Now it’s true that no household has the equivalent of ‘Rosey’ taking care of domestic duties and whatnot, but is there reason to think we might all have our own Rosey in the not-too-distant future?

Could be.

We’re really no longer in the time where A.I. was in its fledgling stage. It’s definitely starting to get wings, and the premise for what’s possible with artificial intelligence is really something. Of course there’s going to be a real need for measures and protocols in place to keep it on the ‘good’ side of things, but man oh man there’s reason to be excited about all the potential here.

And it’s the type of potential that is pretty enticing for anyone who works or provides services in the digital space, and that of course includes us here at 4GoodHosting along with every other Canadian web hosting provider. This kind of stuff always makes the grade with what we’ll consider newsworthy, and so we thought it would be worthwhile to point out major milestones for artificial intelligence that we’ve seen across the year that’s drawing to close now.

Here are six of the main developments and emerging themes seen in artificial intelligence during 2020.

Language Comprehension

Take your average year and a text-generating tool likely won’t make the cut as one of the most exciting new A.I. developments. But this is 2020 (the year the Jetsons were living in if you didn’t get the cartoon reference earlier) and GPT-3 is no ordinary text-generating tool.

For those in the know, it’s the follow-up for GPT-2, at the time labeled as the ‘world’s most dangerous algorithm’ with all the sensationalism you care to attach to that. GPT-3 is a cutting-edge autoregressive natural-language-processing neural network created by the research lab OpenAI. All it needs is the first few lines at the beginning of a news story to then generate impressively accurate text matching the style and content of the initial few lines. It can even generate that quotes that fit as naturally as those you’d get from a breathing interviewee.

Parameters are the connections that are tuned in order to achieve performance. GPT-3 boasts an astonishing 175 billion parameters and is reportedly running up a bill of about $12 million to train.

Microsoft’s Turing Natural Language Generation (T-NLG) can get a shout out here too. After debuting in February 2020 it was a massive 17 billion parameters in size and the largest language model to date. T-NLG is capable of generating the necessary words to complete unfinished sentences, as well as generate direct answers to questions and summarize documents.

Bigger Models

GPT-3 and T-NLG themselves are part of another milestone too. Extensive resource utilization is – as you’d expect – the norm with A.I. development and along with that comes enormous models with huge training costs. Neural networks with upward of a billion parameters are fast becoming the norm.

And more parameters are required all the time.

Look no further than new models like Meena, Turing-NGL, DistilBERT, and BST 9.4B all surpassing 1 billion parameters. More parameters doesn’t necessarily mean better performance in every case. But they do make text-generating tools much more capable with a large range of functions. That’s important, because of real brain-like artificial intelligence is the aim then more parameters are a must.

Philanthropy of Sorts

It’s not only computer scientists who are especially pleased about all of this. Researchers from other disciplines have utilized A.I. to benefit the less fortunate of humanity too. Examples are A.I.-backed technologies diagnosing tinnitus from brain scans or mind-reading headsets that use machine learning to help vocally impaired wearers turn thoughts into words.

The best example here may be DeepMind’s AlphaFold and its contribution to bio sciences. It’s impressively accurate with predicting the shape of proteins based on their sequence, potentially helping develop new more effective therapies rapidly. This is all really good stuff.

Major Advances in Robotics

As intimidating as the thought of them might be, truth is we need major advances in robotics to live more sustainably on the planet. There’s no shortage already of examples of A.I. and robotics carrying out human tasks, but the real advances have been with robotics designed to be able to be trusted with control functions at locations where living or even an extended stay for a human is impractical or impossible.

Robotics really is one area that is set to explode and powered in doing that by A.I.

Deepfakes

Deepfakes aren’t a 2020 invention, but they sure have stolen a lot of attention this year and for good reason. There’s so much wonder that’s natural when you consider what this technology is capable of, and even more so when you think about what’s possible when any deepfake rendering is then given its own A.I. to both look AND act the part.

In July of 2020, researchers from the Center for Advanced Virtuality at the Massachusetts Institute of Technology created a deepfake video that was incredibly real in its appearance and compelling with the way it showed former President Richard Nixon delivering an alternate address about the moon landings. For a man that’s been dead for many years, he sure did look like his old self in the video.

Regulation of A.I. Will be Needed

With power comes the need for responsibility, and that is certainly going to apply to all the A.I.-powered tools we’re going to see over the next few years. Take for example how Police in Detroit wrongly arrested a man after an algorithm mistakenly matched the photo on his drivers’ license with blurry CCTV footage. This then lead to IBM, Amazon, and Microsoft all announcing they would reevaluate the use of their facial-recognition technologies.

This will very much apply to deepfakes too. California passed the AB-730 law designed to criminalize the use of deepfakes and make it clear that those with the technology aren’t to use if for no good.

All this A.I. stuff is really fascinating, and we are as enthused as anyone to see what 2021 has in store along the same lines.

You may also like: