What Devs Should Know About ChatGPT and LLMs with GitHub's Brian Randell
With so much evolving (and occasionally inaccurate) discourse out there around ChatGPT it's critical for devs to learn best practices, tips and guidance from the true experts in these fields.
The past six weeks have been some of the most eventful in the history of artificial intelligence, sending developers, IT pros and end users scrambling to test drive GPT-4, the latest and most sophisticated (so far) incarnation of OpenAI's neural network machine learning model.
With so much new, evolving (and occasionally inaccurate) discourse out there around ChatGPT, generative AI and large language models (LLMs), it's critical for developers to learn best practices, tips and guidance from the true experts in these fields.
Brian Randell, product marketing director at GitHub and longtime Microsoft MVP, has over two decades' worth of experience in software development. Lately, he's had a front-row seat to some of the biggest developments in the AI space since ChatGPT hit the scene. Next month, Randell is sharing his knowledge with developers at the VSLive! conference in Nashville, Tenn. His session, Unveiling the Power of ChatGPT and Large Language Models: A Comprehensive Guide, will, in his own words, "explore the inner workings of ChatGPT and how to effectively utilize its potential." It's part of the brand-new Cutting-Edge AI track at VSLive!, which also includes expert-led sessions on (among other things) what developers need to know about Azure OpenAI, how .NET developers can best use GitHub Copilot, and all the reasons why ChatGPT won't take your job.
Ahead of his May 16 session, we caught up with Randell to get his thoughts on the recent deluge of generative AI developments, including how he stays on top of the news and the exciting new capabilities he thinks are still to come. Make sure to register for VSLive! Nashville to attend his full session and take advantage of the new AI track -- the early-bird deadline is April 14, and regular registrations end on May 15.
VisualStudioMagazine: Is this what we're in for during the immediate future -- one breathtaking AI advance after another after another?
Randell: While it is difficult to predict the exact trajectory of AI development, the recent advancements in AI tools, such as ChatGPT-4 and Copilot X, suggest that we are indeed in a period of rapid innovation in the field. As research continues to push the boundaries of what AI can do, we can expect to see more groundbreaking technologies emerging.
However, it is important to keep in mind that the pace of these breakthroughs may not always be consistent. There might be periods of slower progress as researchers face new challenges or work to refine existing technologies. Additionally, as AI capabilities advance, researchers and developers will need to address ethical, safety and regulatory concerns to ensure that the AI technologies are used responsibly and in the best interests of society.
How do you keep up with it all?
For starters, I have a component of my day job that requires I know relevant pieces. For example, I've been using GitHub Copilot for months before we released it last June for individuals to license. Another benefit is that I read quite quickly and have a good memory. I also rely upon friends and those smarter than me to help me understand. Human interaction is key to being successful in this industry.
"The recent advancements in AI tools, such as ChatGPT-4 and Copilot X, suggest that we are indeed in a period of rapid innovation in the field. As research continues to push the boundaries of what AI can do, we can expect to see more groundbreaking technologies emerging."
Brian Randell, Product Marketing Director, GitHub
What are some especially novel applications that developers have created to put ChatGPT and GPT-4 to use?
I think there's a bunch of cool stuff out there. But I'd like to share my favorites for my talk.
What are some super-easy ways developers can get started with the new tech?
Sign up and try it. Many of the tools are free during pre-release. Even the release cost is often not that expensive if you consider what the tool can do to change your life and make you more productive. In addition, companies like GitHub offer free plans for qualified students.
Have you envisioned any mind-blowing future applications of the tech that we're not quite ready to implement yet?
Let's just say I have to skip that question.
Is it true that data scientists and AI researchers don't fully understand how LLMs work -- even their creators?
Like all the answers, this is my opinion using information I've gleaned from the web.
It seems true that data scientists and AI researchers do not fully understand how large language models (LLMs) like GPT-4 work in every aspect, despite being their creators. LLMs are highly complex and are based on deep learning architectures with millions or billions of parameters. These parameters are adjusted during training, allowing the model to generate outputs based on the patterns it has learned from the data.
I do know that researchers are actively working on improving the interpretability of AI models, which can help build trust and improve the deployment of AI technologies in real-world applications.
When will be able to just talk to our consumer computers and tell them what to do with voice interfaces, instead of just mousing and keyboarding all the time?
Who knows? The rate of change is amazing but if I had that answer, I don't think I'd be providing these responses.
With all these fast-paced advancements coming from rival companies seeking to establish AI supremacy, do you worry about not enough attention being paid to ethical concerns, legal considerations, bias mitigation, safety, responsibility and so on?
Do I worry? No. Why? I can't control it. What I do know is I work for GitHub and, by extension, Microsoft and we're focused on responsible AI. Beyond what I do, I know that there are bad actors that will use AI for "evil." My hope is that more humans will be focused on ethical and responsible AI uses.
What's the main thing you hope attendees come away with from your session?
I hope they're excited to try out the latest tools and technologies. I want them to research and form their own opinions and work to make the world a better place.
About the Author
Gladys Rama (@GladysRama3) is the editor of Redmondmag.com, RCPmag.com and AWSInsider.net, and the editorial director of Converge360.