Exploring the Limitless Potential of ChatGPT in Research

Last month took place one of the crucial necessary conferences for the High Energy Physics (HEP) community in Computing: the so-called CHEP 2023, standing for Computing on High Energy Physics and Nuclear Physics — yes, easy! 🙂

As a Computer Engineer working at CERN, it’s a serious event: it’s the chance to see the trend of the most recent technologies in our field. Nevertheless, although I used to be fully aware of the present popularity of ChatGPT, I used to be not expecting to seek out any talks on the topic. But I used to be totally fallacious, indeed there have been a pair!

I discovered them very appealing, so in this text, I would love to depict the primary take home-messages of such talks. ChatGPT just isn’t only reshaping our every day tasks but in addition major research areas corresponding to the HEP one.

Let’s explore what’s coming!

The HEP community refers back to the global network of scientists, researchers, engineers, technicians, and institutions involved in the sphere of High Energy Physics. This community is devoted to the study of the fundamental constituents of matter, the forces that govern their interactions, and the exploration of the fundamental laws of the universe.

CHEP is a series of conferences that deal with using computing, software, and data management in the sphere of HEP — and Nuclear Physics too.

Actually, CHEP is sort of old conference. The primary one took place in 1985, and since then, it has been organized biennially. Overall, CHEP conferences play an important role in driving advancements in computing and data management.

CHEP serves as a platform for knowledge exchange, collaboration, and the exploration of latest computing techniques. That’s the reason I used to be actually that surprised: if something appears at CHEP, it’s more likely to be an incoming trend! And on this last CHEP 2023, we had two plenary sessions about ChatGPT in HEP.

Ready?

The primary plenary session about ChatGPT got here very early within the schedule by David Dean from the Jefferson Lab. Titled Evolution and Revolutions in Computing: Science on the Frontier, David provided an unlimited overview of the most recent revolutions in computing. And fear not, ChatGPT was one among them!

He concretely targeted the query around if ChatGPT can do physics, and the message was clear: it’s a mind-blowing tool that may pass physics exams too, but there may be a serious flaw which will stop ChatGPT to be incorporated as a tool within the near future: Model Hallucinations.

Model Hallucinations

Despite the model’s capabilities to retrieve human-like responses, there are moments where it maintains an inclination to make up facts, to double down on misinformation, and to perform tasks incorrectly. Those incorrect responses are generally known as hallucinations.

In truth, giving incorrect answers just isn’t the issue itself. The primary issue is that ChatGPT often exhibits these tendencies in a convincing and authoritative manner. Hallucinations are sometimes even present in the shape of highly detailed information, giving a fallacious sense of accuracy to the reader, and increasing the danger of overreliance. And that is certainly an issue within the research community.

With a purpose to use ChatGPT as a trustful helper tool, hallucinations must be controlled. Currently, ChatGPT will try to offer a solution to any of the given queries, even when it has not enough information concerning the goal topic.

There ought to be nothing bad about ChatGPT admitting it just isn’t in a position to provide an accurate response to a given prompt and it will make the tool more suitable in accurate environments corresponding to within the HEP research.

The second plenary session touching on ChatGPT was entitled Radically different futures for HEP enabled by AI/ML, given by Kyle Crammer from the University of Wisconsin-Madison.

This second talk was more optimistic concerning the introduction of ChatGPT as a helpful asset within the HEP toolkit. In truth, Kyle referenced one other talk from Christian Weber from the Brookhaven National Laboratory through which he presented real use cases of ChatGPT as a coding assistant, especially to migrate and convert code to recent platforms. In truth, ChatGPT already implements a Python interpreter for coding purposes.

Each experiment within the HEP community has its own coding templates, i.e., although coding in Python, scientists must keep on with some classes or style conventions. One in all the use cases presented was to fine-tune ChatGPT to jot down evaluation code based on the experiment template.

Attracted by this use-case, I attempted to generate a template for an evaluation on my current experiment, the CMS Experiment at CERN, Switzerland, and ChatGPT perfectly generated a primary template. And I used to be simply using the online interface, imagine how powerful it might be after fine-tuning it with relevant data.

In response to the presentation, even when sometimes the evaluation was not accurate enough, it allowed for generating a primary template or backbone for the evaluation. This concept was explored to offer faster onboarding for brand spanking new experiment members and construct prototypes faster, amongst other use-cases.

We cannot deny that Large Language Models (LLMs) corresponding to ChatGPT are changing our method to seek for information, construct applications, and even coding.

As with every advancement in technology, I feel it is cheap to guage any recent tool with the intention to leverage its advantages and apply them to our primary fields. These two plenary sessions are only two examples of this evaluation process in a giant research community corresponding to the HEP one.

While some evaluations may discard ChatGPT in the intervening time as a research helper, others may allow the incorporation of such tools in concrete and delimited domains. In any case, I imagine it’s important to not fear AI and proceed to evolve with it, analyzing its advantages, knowing the way to optimize its performance on your goal domain, and rather more importantly, being aware of the issues to maintain the critical spirit at all times alert!