If AI has anything to say to 2020, its you cant touch this.
Last year may have severed our connections with the physical world, but in the digital realm, AI thrived. Take NeurIps, the crown jewel of AI conferences. While lacking the usual backdrop of the dazzling mountains of British Columbia or the beaches of Barcelona, the annual AI extravaganza highlighted a slew of big picture problemsbias, robustness, generalizationthat will encompass the field for years to come.
On the nerdier side, scientists further explored the intersection between AI and our own bodies. Core concepts in deep learning, such as backpropagation, were considered a plausible means by which our brains assign fault in biological networksallowing the brain to learn. Others argued its high time to double-team intelligence, combining the reigning AI golden child methoddeep learningwith other methods, such as those that guide efficient search.
Here are four areas were keeping our eyes on in 2021. They touch upon outstanding AI problems, such as reducing energy consumption, nixing the need for exuberant learning examples, and teaching AI some good ole common sense.
Youve heard this a billion times: deep learning is extremely greedy, in that the algorithms need thousands (if not more) examples to showcase basic signs of learning, such as identifying a dog or a cat, or making Netflix or Amazon recommendations.
Its extremely time-consuming, wasteful in energy, and a head-scratcher in that it doesnt match our human experience of learning. Toddlers need to see just a few examples of something before they remember it for life. Take the concept of dogregardless of the breed, a kid whos seen a few dogs can recognize a slew of different breeds without ever having laid eyes on them. Now take something completely alien: a unicorn. A kid who understands the concept of a horse and a narwhal can infer what a unicorn looks like by combining the two.
In AI speak, this is less than one-shot learning, a sort of holy-grail-like ability that allows an algorithm to learn more objects than the amount of examples it was trained on. If realized, the implications would be huge. Currently-bulky algorithms could potentially run smoothly on mobile devices with lower processing capabilities. Any sort of inference, even if it doesnt come with true understanding, could make self-driving cars far more efficient at navigating our object-filled world.
Last year, one team from Canada suggested the goal isnt a pipe dream. Building on work from MIT analyzing hand-written digitsa common toy problem in computer visionthey distilled 60,000 images into 5 using a concept called soft labels. Rather than specifying what each number should look like, they labeled each digitsay, a 3as a percentage of 3, or 8, or 0. Shockingly, the team found that with carefully-constructed labels, just two examples could in theory encode thousands of different objects. Karen Hao at MIT Technology Review gets into more detail here.
For everything AI can do, its flawed at defending insidious attacks targeting its input data. Slight or seemingly random perturbations to a datasetoften undetectable by the human eyecan enormously alter the final output, something dubbed brittle for an algorithm. Too abstract? An AI trained to recognize cancer from a slew of medical scans, annotated in yellow marker by a human doctor, could learn to associate yellow with cancer. A more malicious example is nefarious tampering. Stickers placed on a roadway can trick Teslas Autopilot system to mistake lanes and careen into oncoming traffic.
Brittleness requires AI to learn a certain level of flexibility, but sabotageor adversarial attacksis becoming an increasingly recognized problem. Here, hackers can change the AIs decision-making process with carefully-crafted inputs. When it comes to network security, medical diagnoses, or other high-stakes usage, building defense systems against these attacks is critical.
This year, a team from the University of Illinois proposed a powerful way to make deep learning systems more resilient. They used an iterative approach, having two neural nets battle it outone for image recognition, and the other for generating adversarial attacks. Like a cat-and-mouse game, the enemy neural net tries to fool the computer vision network into recognizing things that are fictitious; the latter network fights back. While far from perfect, the study highlights one increasingly popular approach to make AI more resilient and trustworthy.
One of the most impressive algorithms this year is GPT-3, a marvel by OpenAI that spits out eerily human-like language. Dubbed one of the most interesting and important AI systems ever produced, GPT-3 is the third generation of an algorithm that produces writing so natural that at a glance its hard to decipher machine from human.
Yet GPT-3s language proficiency is, upon deeper inspection, just a thin veil of intelligence. Because its trained on human language, its also locked into the intricacies and limitations of our everyday phraseswithout any understanding of what they mean in the real world. Its akin to learning slang from Urban Dictionary instead of living it. An AI may learn to associate rain with cats and dogs in all situations, gaining its inference from the common vernacular describing massive downpours.
One way to make GPT-3 or any natural language-producing AI smarter is to combine it with computer vision. Teaching language models to see is an increasingly popular area in AI research. The technique combines the strength of language with images. AI language models, including GPT-3, learn through a process called unsupervised training, which means they can parse patterns in data without explicit labels. In other words, they dont need a human to tell them grammatical rules or how words relate to one another, which makes it easier to scale any learning by bombarding the AI with tons of example text. Image models, on the other hand, better reflect our actual reality. However, these require manual labeling, which makes the process slower and more tedious.
Combining the two yields the best of both worlds. A robot that can see the world captures a sort of physicalityor common sensethats missing from analyzing language alone. One study in 2020 smartly combined both approaches. They started with language, using a scalable approach to write captions for images based on the inner workings of GPT-3 (details here). The takeaway is that the team was able to connect the physical worldrepresented through imagesby linking it with language on how we describe the world.
Translation? A blind, deaf, and utterly quarantined AI learns a sort of common sense. For example, cats and dogs can just mean pets, rather than rain.
The trick is still mostly experimental, but its an example of thinking outside the artificial confines of a particular AI domain. By combining the two areasnatural language processing and computer visionit works better. Imagine an Alexa with common sense.
Speaking of thinking outside the box, DeepMind is among those experimenting with combining different approaches to AI into something more powerful. Take MuZero, an Atari-smashing algorithm they released just before Christmas.
Unlike DeepMinds original Go, poker, and chess-slaying AI wizard, MuZero has another trick up its sleeve. It listens to no one, in that the AI doesnt start with previous knowledge of the game or decision-making processes. Rather, it learns without a rulebook, instead observing the games environmentakin to a novice human observing a new game. In this way, after millions of games, it doesnt just learn the rules, but also a more general concept of policies that could lead it to get ahead and evaluate its own mistakes in hindsight.
Sounds pretty human, eh? In AI vernacular, the engineers combined two different approaches, decision trees and a learned model, to make an AI great at planning winning moves. For now, its only been shown to master games at a level similar to AlphaGo. But we cant wait to see what this sort of cross-fertilization of ideas in AI can lead to in 2021.
Image Credit: Oleg Gamulinskiy from Pixabay
Read the original:
- How to get the most from ChessBase 16 - Chessbase News - January 20th, 2021
- What are the key factors and impact of interventions in Englands COVID-19 epidemic? - News-Medical.Net - January 20th, 2021
- Richmond Hill basketball coaches piece together scouting reports on opponents - Savannah Morning News - January 20th, 2021
- What The Biden Presidency Means For Privacy, Consumer Protection And Antitrust - AdExchanger - January 20th, 2021
- An intro to the fast-paced world of artificial intelligence - MIT News - January 20th, 2021
- Counter-UAS System Market Size, Growth Factor,Analysis And Forecast By 2027| Lockheed Martin, Thales, Boeing - NeighborWebSJ - January 20th, 2021
- Anglo Singapore International School are delighted to welcome you for a school visit. - Bangkok Post - January 20th, 2021
- A Guide to Optimizing Your Mental Health in 2021 - Dan's Papers - January 20th, 2021
- Chess Global Market : Industry Analysis and Forecast (2019-2026) by Type, Application, and Region - The Courier - January 15th, 2021
- ONTARIO: How one womans online chess school is getting young girls worldwide excited about the game - TimminsToday - January 15th, 2021
- LaFleur Won't Be Able to Fool McVay on Saturday - Zone Coverage - January 15th, 2021
- Lewis Carroll Death Anniversary: Here are Some of the Author's Greatest Works - News18 - January 15th, 2021
- Reflecting on 'The Queen's Gambit': Are women genetically hardwired to underperform men in chess? - Genetic Literacy Project - January 13th, 2021
- Chess Market: Global Strategies and Insight driven transformation 2020 2026 - SoccerNurds - January 13th, 2021
- Why Its Not Too Late to Learn New Skills - The New Yorker - January 13th, 2021
- Global Counter UAV Market 2020 Trends, Demand and Scope with Outlook, Business Strategies and Forecast 2025 - SoccerNurds - January 13th, 2021
- The Seedy Chess World of Queens Gambit Is the Shows Masterstroke - Observer - January 9th, 2021
- Chess Market Study by Global and Regional Analysis with Forecast | The House of Staunton, ChessSUA, CNCHESS - NeighborWebSJ - January 9th, 2021
- How this Ontario woman's online chess school is getting young girls worldwide excited about the game - Toronto Star - January 9th, 2021
- Americans are living 30 years longer on average and that's a big plus for businesses - Palo Alto Online - January 9th, 2021
- From Text to Tablet: How to Learn in Lockdown - Global Citizen - January 9th, 2021
- Overcoming lockdowns and uncertainty to triumph at the end of a tough year - Neos Kosmos - January 9th, 2021
- Science-backed strategies to deal with stress and anxiety - World Economic Forum - January 9th, 2021
- 'The Queen's Gambit' Is About More Than Just Chess - Study Breaks - January 6th, 2021
- Barnegat's Kasey Keshecki is our Student of the Week - wobm.com - January 6th, 2021
- Film Study: The Ohio State Offense Was Prepared to Peak in the College Football Playoff | Eleven Warriors - Eleven Warriors - January 6th, 2021
- Chess Market Statistical Overview 2021-2026 with Industry Insights and Growth by ChessSUA, CNCHESS, ChessBaron, Chessncrafts, Chessbazaar.com, and... - January 6th, 2021
- Global Used Construction Machinery Market Precise Scenario Covering Trends, Competitive Scenario, Business Opportunities and Growth Forecast during... - January 6th, 2021
- Anti-UAV Defense System Market Analysis, Opportunities, Competition Scenario, Threats, Future Prospects, Trends, Applications, And Growth Forecast To... - January 6th, 2021
- Global Chess Market to 2026 Industry Perspective, Comprehensive Analysis, and Forecast - Factory Gate - January 4th, 2021
- From the shadows into the light the man who broke the UFO embargo grew up in Sarasota - Sarasota Herald-Tribune - January 4th, 2021
- Six ways to reboot your brain after a hard year of Covid-19 according to science - iNews - January 4th, 2021
- Neuroscientists explain six ways to reboot your brain after a hard year of COVID-19 - PsyPost - January 4th, 2021
- Nithiin starrer Check first glimpse will leave you intrigued about his character | WATCH - Republic World - January 4th, 2021
- Good News Stories That Actually Made Us Smile in 2020 - Newsweek - January 4th, 2021
- Irelands modern druids: There is hidden knowledge you are asked not to share - The Irish Times - January 4th, 2021
- Keep Calm and QAnon - Slate - December 31st, 2020
- Beginners by Tom Vanderbilt review it's never too late to learn - The Guardian - December 31st, 2020
- Yes, 2020 was a historically hellish year. Now for the good news | TheHill - The Hill - December 31st, 2020
- Literature alights that was once burned books | Dec. 30 - Jan. 5, 2020 - Real Change News - December 31st, 2020
- Google DeepMind unveils AI that can master chess without even knowing the rules - The Independent - December 29th, 2020
- Coronavirus: Latest updates on COVID-19 crisis around the world - Outlook India - December 29th, 2020
- Chess Market 2020 Future Growth Analysis With Focusing Key Players | The House of Staunton, ChessSUA, CNCHESS, ChessBaron - Factory Gate - December 29th, 2020
- Counter-UAV (C-UAV) Systems Market Research Report (2020-2026): Key Trends and Opportunities |Lockheed Martin, Thales, Boeing - Factory Gate - December 29th, 2020
- Kwanzaa in Louisville on heels of social justice movement - Alton Telegraph - December 29th, 2020
- The New China Challenge Stems from Beijings Old Ambitions - The National Interest - December 29th, 2020
- International Chess Market Opportunities And Forecast Analysis up to 2026 | The House of Staunton, ChessSUA, CNCHESS - The Monitor - December 29th, 2020
- WHO: Learn from COVID Pandemic, It Wont be the Last - Voice of America - December 27th, 2020
- How 'The Flight Attendant' and 'The Queen's Gambit' Are the Same - GQ - December 27th, 2020
- Online shopping is here to stay in 2021: Refinitiv - Yahoo Money - December 27th, 2020
- Many Jews refrain from learning Torah on Christmas Eve and call it Nittel Nacht. Here's my family tradition to spread light. - JTA News - Jewish... - December 26th, 2020
- Global Chess Market Demand, Revenue, Trends, Profit Analysis and Landscape Outlook to 2026|| The House of Staunton, ChessSUA, CNCHESS, ChessBaron -... - December 26th, 2020
- The kings gambit: Viswanathan Anand enters a different phase of his career - Sportstar - December 26th, 2020
- 19 Calm Coffee Places In Beirut Where You Can Focus While Working - The961 - December 26th, 2020
- DeepMinds Latest AI Can Win Games Without Knowing the Rules - Beebom - December 26th, 2020
- Chess Table Market Size By Type, By Application, By Geography, By Top Companies And Forecast To 2027 - LionLowdown - December 26th, 2020
- Impact Of Covid 19 On Counter UAV Industry 2020 Market Challenges, Business Overview And Forecast Research Study 2026 - Factory Gate - December 23rd, 2020
- Bellamy Richardson The Williams Record - The Williams record - December 23rd, 2020
- Between the sacred and the secular - New Statesman - December 23rd, 2020
- In a winter of change, Eagles are learning that Jalen Hurts is no snowflake - NJ.com - December 22nd, 2020
- PhD grad uses math to explore the brain - ASU Now - December 22nd, 2020
- OPINION: Sometimes it's good to have a team - Arkansas Online - December 22nd, 2020
- Business is continuing its cloud migration. Train to become an Azure and AWS tech professional - The Next Web - December 22nd, 2020
- Staying On Top: How Chess Masters Study The Game ... - December 20th, 2020
- Struggling local entertainment venues: Its time for the state to show it CARES - Buffalo News - December 20th, 2020
- The best series and shows on Netflix in 2020 - RNZ - December 20th, 2020
- 21 Ways to Greater Well-Being in 2021 - Newswise - December 20th, 2020
- The Best Video Games of 2020 - The New Yorker - December 18th, 2020
- Memo to Blinken: Protect the brains of State Department employees - Federal News Network - December 18th, 2020
- Chadwick Boseman Gives His Final Performance in 'Ma Raineys Black Bottom', and It's Gut-Wrenching - Glamour - December 18th, 2020
- Chess Course for High Performance: Grandmaster Package - December 17th, 2020
- Reasons to study international studies and diplomacy - Modern Diplomacy - December 17th, 2020
- Global Chess Market Demand and Opportunities 2026|| Break Down by Top Companies, Countries, Applications, Challenges, Opportunities and Forecast -... - December 16th, 2020
- What is a 'Flow State' and How Can MTBers Get Into It More Often? - Singletracks.com - December 16th, 2020
- This is why you make bad decisions on the golf course (and how to fix it) - Golf.com - December 16th, 2020
- Author Sophie Cousens talks about her book, 'This Time Next Year' - Yahoo! Voices - December 16th, 2020
- The board games market by revenue is expected to grow at a CAGR of approx. 13% during the period 20212026 - GlobeNewswire - December 16th, 2020
- The Queen's Gambit: 10 Times Beth Is The Heroine Fans Needed - Screen Rant - December 15th, 2020
- Chelmsford High School October Students of the Month - Wicked Local - December 15th, 2020
- Struggling Museums Are Increasingly Relying on the Generosity of Artists to Convince Private Donors to Bail Them Out - artnet News - December 15th, 2020