Robot Technology News
ROBO SPACE
Solving a machine-learning mystery
MIT researchers found that massive neural network models that are similar to large language models are capable of containing smaller linear models inside their hidden layers, which the large models could train to complete a new task using simple learning algorithms. Credits:Image: Jose-Luis Olivares, MIT
Solving a machine-learning mystery
by Adam Zewe for MIT News
Boston MA (SPX) Feb 08, 2023

Large language models like OpenAI's GPT-3 are massive neural networks that can generate human-like text, from poetry to programming code. Trained using troves of internet data, these machine-learning models take a small bit of input text and then predict the text that is likely to come next.

But that's not all these models can do. Researchers are exploring a curious phenomenon known as in-context learning, in which a large language model learns to accomplish a task after seeing only a few examples - despite the fact that it wasn't trained for that task. For instance, someone could feed the model several example sentences and their sentiments (positive or negative), then prompt it with a new sentence, and the model can give the correct sentiment.

Typically, a machine-learning model like GPT-3 would need to be retrained with new data for this new task. During this training process, the model updates its parameters as it processes new information to learn the task. But with in-context learning, the model's parameters aren't updated, so it seems like the model learns a new task without learning anything at all.

Scientists from MIT, Google Research, and Stanford University are striving to unravel this mystery. They studied models that are very similar to large language models to see how they can learn without updating parameters.

The researchers' theoretical results show that these massive neural network models are capable of containing smaller, simpler linear models buried inside them. The large model could then implement a simple learning algorithm to train this smaller, linear model to complete a new task, using only information already contained within the larger model. Its parameters remain fixed.

An important step toward understanding the mechanisms behind in-context learning, this research opens the door to more exploration around the learning algorithms these large models can implement, says Ekin Akyurek, a computer science graduate student and lead author of a paper exploring this phenomenon. With a better understanding of in-context learning, researchers could enable models to complete new tasks without the need for costly retraining.

"Usually, if you want to fine-tune these models, you need to collect domain-specific data and do some complex engineering. But now we can just feed it an input, five examples, and it accomplishes what we want. So, in-context learning is an unreasonably efficient learning phenomenon that needs to be understood," Akyurek says.

Joining Akyurek on the paper are Dale Schuurmans, a research scientist at Google Brain and professor of computing science at the University of Alberta; as well as senior authors Jacob Andreas, the X Consortium Assistant Professor in the MIT Department of Electrical Engineering and Computer Science and a member of the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL); Tengyu Ma, an assistant professor of computer science and statistics at Stanford; and Danny Zhou, principal scientist and research director at Google Brain. The research will be presented at the International Conference on Learning Representations.

A model within a model
In the machine-learning research community, many scientists have come to believe that large language models can perform in-context learning because of how they are trained, Akyurek says.

For instance, GPT-3 has hundreds of billions of parameters and was trained by reading huge swaths of text on the internet, from Wikipedia articles to Reddit posts. So, when someone shows the model examples of a new task, it has likely already seen something very similar because its training dataset included text from billions of websites. It repeats patterns it has seen during training, rather than learning to perform new tasks.

Akyurek hypothesized that in-context learners aren't just matching previously seen patterns, but instead are actually learning to perform new tasks. He and others had experimented by giving these models prompts using synthetic data, which they could not have seen anywhere before, and found that the models could still learn from just a few examples. Akyurek and his colleagues thought that perhaps these neural network models have smaller machine-learning models inside them that the models can train to complete a new task.

"That could explain almost all of the learning phenomena that we have seen with these large models," he says.

To test this hypothesis, the researchers used a neural network model called a transformer, which has the same architecture as GPT-3, but had been specifically trained for in-context learning.

By exploring this transformer's architecture, they theoretically proved that it can write a linear model within its hidden states. A neural network is composed of many layers of interconnected nodes that process data. The hidden states are the layers between the input and output layers.

Their mathematical evaluations show that this linear model is written somewhere in the earliest layers of the transformer. The transformer can then update the linear model by implementing simple learning algorithms.

In essence, the model simulates and trains a smaller version of itself.

Probing hidden layers
The researchers explored this hypothesis using probing experiments, where they looked in the transformer's hidden layers to try and recover a certain quantity.

"In this case, we tried to recover the actual solution to the linear model, and we could show that the parameter is written in the hidden states. This means the linear model is in there somewhere," he says.

Building off this theoretical work, the researchers may be able to enable a transformer to perform in-context learning by adding just two layers to the neural network. There are still many technical details to work out before that would be possible, Akyurek cautions, but it could help engineers create models that can complete new tasks without the need for retraining with new data.

"The paper sheds light on one of the most remarkable properties of modern large language models - their ability to learn from data given in their inputs, without explicit training. Using the simplified case of linear regression, the authors show theoretically how models can implement standard learning algorithms while reading their input, and empirically which learning algorithms best match their observed behavior," says Mike Lewis, a research scientist at Facebook AI Research who was not involved with this work.

"These results are a stepping stone to understanding how models can learn more complex tasks, and will help researchers design better training methods for language models to further improve their performance."

Moving forward, Akyurek plans to continue exploring in-context learning with functions that are more complex than the linear models they studied in this work. They could also apply these experiments to large language models to see whether their behaviors are also described by simple learning algorithms. In addition, he wants to dig deeper into the types of pretraining data that can enable in-context learning.

"With this work, people can now visualize how these models can learn from exemplars. So, my hope is that it changes some people's views about in-context learning," Akyurek says. "These models are not as dumb as people think. They don't just memorize these tasks. They can learn new tasks, and we have shown how that can be done."

Research Report:"What Learning Algorithm Is In-Context Learning? Investigations with Linear Models"

Related Links
Computer Science and Artificial Intelligence Laboratory
All about the robots on Earth and beyond!

Subscribe Free To Our Daily Newsletters
Tweet

RELATED CONTENT
The following news reports may link to other Space Media Network websites.
ROBO SPACE
Japan rolls out 'humble and lovable' delivery robots
Fujisawa, Japan (AFP) Feb 8, 2023
"Excuse me, coming through," a four-wheeled robot chirps as it dodges pedestrians on a street outside Tokyo, part of an experiment businesses hope will tackle labour shortages and rural isolation. From April, revised traffic laws will allow self-driving delivery robots to navigate streets across Japan. Proponents hope the machines could eventually help elderly people in depopulated rural areas get access to goods, while also addressing a shortage of delivery workers in a country with chronic lab ... read more

ROBO SPACE
What we know about mysterious objects downed by US

Speculation grows as US shoots down new mystery object

US has had contacts with China over 'spy balloon': Pentagon

US adds six China entities tied to balloon program to blacklist

ROBO SPACE
High efficiency mid- and long-wave optical parametric oscillator pump source and its applications

Automating the math for decision-making under uncertainty

Understanding laser accelerated electron radiation through terahertz emissions

Turkey's once mighty developers under fire after quake

ROBO SPACE
Spinning up a 'flip-flop' qubit

Atom-thin walls could smash size, memory barriers in next-gen devices

Developing practical quantum computers that can solve big challenges of our time

Researchers pioneer process to stack micro-LEDs

ROBO SPACE
UN nuclear chief Grossi due in Russia for Ukraine talks

Fermi Energia chooses GE Hitachi BWRX-300 Small Modular Reactor for Estonia

Belgium plans to extend life of three nuclear reactors

Belgium to shut down controversial nuclear reactor

ROBO SPACE
Guantanamo 'high value' prisoner released to Belize

Sweden moves to toughen anti-terror laws

British Army serviceman in court for terror offence

Myanmar military accused of war crimes in German criminal complaint

ROBO SPACE
All who can should pay even for their basic greenhouse gas emissions

S.Africa mining and energy giants thwarting climate goals: study

Energy industry must be part of climate fight, says COP president

France urges 'transparency' over US climate subsidies

ROBO SPACE
Controllable 'defects' improve performance of lithium-ion batteries

Hungary protests against Chinese battery plant defy 'Orbanomics'

Ford to build new US electric battery plant with Chinese partner

AiDash launches joint grid resilience offering with Schneider Electric

ROBO SPACE
China's Deep Space Exploration Lab eyes top global talents

Chinese astronauts send Spring Festival greetings from space station

China to launch 200-plus spacecraft in 2023

China's space industry hits new heights

Subscribe Free To Our Daily Newsletters




The content herein, unless otherwise known to be public domain, are Copyright 1995-2024 - Space Media Network. All websites are published in Australia and are solely subject to Australian law and governed by Fair Use principals for news reporting and research purposes. AFP, UPI and IANS news wire stories are copyright Agence France-Presse, United Press International and Indo-Asia News Service. ESA news reports are copyright European Space Agency. All NASA sourced material is public domain. Additional copyrights may apply in whole or part to other bona fide parties. All articles labeled "by Staff Writers" include reports supplied to Space Media Network by industry news wires, PR agencies, corporate press officers and the like. Such articles are individually curated and edited by Space Media Network staff on the basis of the report's information value to our industry and professional readership. Advertising does not imply endorsement, agreement or approval of any opinions, statements or information provided by Space Media Network on any Web page published or hosted by Space Media Network. General Data Protection Regulation (GDPR) Statement Our advertisers use various cookies and the like to deliver the best ad banner available at one time. All network advertising suppliers have GDPR policies (Legitimate Interest) that conform with EU regulations for data collection. By using our websites you consent to cookie based advertising. If you do not agree with this then you must stop using the websites from May 25, 2018. Privacy Statement. Additional information can be found here at About Us.