Emergent abilities in AI: large language models

The public release of ChatGPT was a landmark that surprised many people, both in the general public and researchers working in Artificial Intelligence. All of a sudden it seemed Large Language Models had capabilities that some thought were a decade away or even not possible. It is like the field underwent a "phase transition." This idea turns out to be more than just a physics metaphor. It has been made concrete and rigorous in the following paper.

Emergent Abilities of Large Language Models

Jason Wei, Yi Tay, Rishi Bommasani, Colin Raffel, Barret Zoph, Sebastian Borgeaud, Dani Yogatama, Maarten Bosma, Denny Zhou, Donald Metzler, Ed H. Chi, Tatsunori Hashimoto, Oriol Vinyals, Percy Liang, Jeff Dean, William Fedus

They use the following definition, "Emergence is when quantitative changes in a system result in qualitative changes in behavior," citing Phil Anderson's classic "More is Different" article. [Even though the article does not contain the word emergence]. 

In this paper, we will consider a focused definition of emergent abilities of large language models: 

An ability is emergent if it is not present in smaller models but is present in larger models.

How does one define the "size" or "scale" of a model? Wei et al., note that "Today’s language models have been scaled primarily along three factors: amount of computation, number of model parameters, and training dataset size."

The essence of the analysis in the paper is summarised as follows.

We first discuss emergent abilities in the prompting paradigm, as popularized by GPT-3 (Brown et al., 2020). In prompting, a pre-trained language model is given a prompt (e.g. a natural language instruction) of a task and completes the response without any further training or gradient updates to its parameters.

An example of a prompt is shown below 

Brown et al. (2020) proposed few-shot prompting, which includes a few input-output examples in the model’s context (input) as a preamble before asking the model to perform the task for an unseen inference-time example. 

 The ability to perform a task via few-shot prompting is emergent when a model has random performance until a certain scale, after which performance increases to well-above random. 

An example is shown in the Figure below. The horizontal axis is the number of training FLOPs for the model, a measure of model scale. The vertical axis measures the accuracy of the model to perform a task, Modular Arithmetic, for which the model was not designed, but just given two-shot prompting. The red dashed line is the performance for a random model. The purple data is for GPT-3 and the blue for LaMDA. Note how once the model scale reaches about 10^22 there is a rapid onset of ability.

The Figure below summarises recent results from a range of research groups studying five different language model families. It shows eight different emergent abilities.

Wei et al., point out that "there are currently few compelling explanations for why such abilities emerge the way that they do".

The authors have encountered some common characteristics of emergent properties. They are hard to predict or anticipate before they are observed. They are often universal, i.e., they can occur in a wide range of different systems and are not particularly sensitive to the details of the components. Even after emergent properties are observed, it is still hard to explain why they occur, even when one has a good understanding of the properties of the system at a smaller scale. Superconductivity was observed in 1911 and only explained in 1957 by the BCS theory.

On the positive side, this paper presents hope that computational science and technology are at the point where AI may produce more exciting capabilities. On the negative side, there is also the possibility of significant societal risks such as having unanticipated power to create and disseminate false information, bias, and toxicity.

Aside: One thing I found surprising is that the authors did not reference John Holland, a computer scientist, and his book, Emergence.

I thank Gerard Milburn for bringing the paper to my attention.

Comments

Post a Comment

Popular posts from this blog

What is Herzberg-Teller coupling?

Image
Is it something to do with breakdown of the Born-Oppenheimer approximation? In molecular spectroscopy you occasionally hear this term thrown around. Google scholar yields more than 3000 hits. But I have found its precise meaning and the relevant physics hard to pin down. Quantum mechanics in chemistry by Schatz and Ratner is an excellent book , but the discussion on page 204 did not help me. "Herzberg-Teller" never appears in Atkins' Molecular quantum mechanics. So here is my limited understanding. Herzberg and Teller wanted to understand why one observed certain vibronic (combined electronic and vibrational) transitions that were not expected, particularly some that were expected to be forbidden on symmetry grounds. " Intensity borrowing " occurred. Herzberg and Teller pointed out that his could be understood if the dipole transition moment for the electronic transition depended on the nuclear co-ordinate associated with the vibration. In the Franck-Condo
Read more

Is it an Unidentified Superconducting Object (USO)?

If you look on the arXiv and in Nature journals there is a continuing stream of people claiming to observe superconductivity in some new material. There is a long history of this and it is worth considering the wise observations of Robert Cava , back in 1997, contained in a tutorial lecture. It would have been useful indeed in the early days of the field [cuprate superconductors] to have set up a "commission" to set some minimum standard of data quality and reproducibility for reporting new superconductors. An almost countless number of "false alarms" have been reported in the past decade, some truly spectacular. Koichi Kitazawa from the University of Tokyo coined these reports "USOs", for Unidentified Superconducting Objects , in a clever cross-cultural double entendre likening them to UFOs (Unidentified Flying Objects, which certainly are their equivalent in many ways) and to "lies" in the Japanese translation of USO.   These have caused g
Read more

What should be the order of authors on a conference poster or talk?

I welcome discussion on this point. I don't think it is as sensitive or as important a topic as the author order on papers. With regard to paper authorship my general rule is that the person who does the bulk of the work, including actually writing the paper should be the first author. Doug Natelson has a good post on co-authorship , that I largely agree with. My only difference is that I am not really convinced that good practice prevails in the majority of circumstances. I fear there are increasing numbers of co-authors, particularly senior ones, with marginal contributions. But, what about conference talks and posters? Many of these are based on work that is already or about to be published. Should the author order be identical as the associated papers? I am not sure it should necessarily be. My tentative view is that the person who writes and submits the abstract and actually prepares and presents the post/talk should be the first author. Perhaps they should also highlig
Read more