Tips on the basics of artificial intelligence: Lessons from Jonathan Soma

AI Basics


“This is something that people think is very technical and very math-intensive, and you have to be a good programmer or a computer science expert to understand it all. . But that's not true. Anyone can do it.'' ” can be said to be a person. He's polite, a bit enthusiastic about everything to do with data journalism, and makes understanding AI a walk in the sun.

It was a particularly sunny day when we met at the iMEdD International Journalism Forum 2023. He had already spent an entire day talking about this recent technology, participating as part of a panel on “AI in Newsrooms: A Debate on the Role of AI in Media,” and also speaking. Workshop on “Creating and Breaking AI in Newsrooms.” In the garden of Piraeus 260, he shared with iMEdD how anyone can get started with AI.

overcome the confusion

Knowing the basics means knowing what you have to deal with, but as Soma says, this isn't always easy. “The more we talk about AI and have conversations with people, the more we come to the conclusion that everyone has a different definition of AI, even computer science people.” As Soma says, this It is a system that allows computers to do 'advanced things', such as performing tasks that require human intelligence, often described as evolution. But now computers can learn on their own. As for these seemingly complex definitions, his IBM YouTube video on AI breaks down the basic terms through simple explanations.

With the arrival of ChatGPT in late 2022, understanding how generative AI works has become more complex, but important in the AI ​​conversation. ChatGPT makes technology accessible to all users, all news organizations, and all journalists. “Generative AI is when computers create things,” Soma explains. “Historically, machine learning in AI has produced small things, maybe just a few words or even categories, but generative AI will produce things that are commonly associated with human creativity.” A bit magical, but primarily based on logistics, generative power extends from outputs such as images and videos to code. But its greatness often comes from the ease of language.

This is where generative AI chatbots like ChatGPT come into play.

Jonathan Soma holds a workshop "Artificial Intelligence and Journalism: Making and Breaking AI in Newsrooms". He stands between his two screens with notes and holds a microphone. He is wearing a plaid shirt. The audience appears in the foreground, out of focus.Jonathan Soma holds a workshop "Artificial Intelligence and Journalism: Making and Breaking AI in Newsrooms". He stands between his two screens with notes and holds a microphone. He is wearing a plaid shirt. The audience appears in the foreground, out of focus.
Jonathan Soma during the “Create and Destroy AI in Newsrooms” workshop Photo credit: Alex Grymanis

Chat with GPT

“The first step always starts with ChatGPT, because you can have a conversation and go back and forth.” The Financial Times explains how large-scale language models and transformers – the fundamental mechanisms of generative chatbot technology – mimic human abilities Level's sophisticated text, images, and computer code in the article “Generative AI exists thanks to Transformers.'' So, starting a conversation with a chatbot will help you understand how the AI ​​tool works, how to interact with it, and what information it needs. This is a great way to get introduced to technology and explore unlimited user options. Try “feeding” one of your stories to write a summary, write a lead, or ask about a general subject to provide a possible new angle.

In February 2023, Reuters wrote that OpenAI's ChatGPT had the second fastest growing user base. ChatGPT 3.5 is free, but the latest update, his ChatGPT 4, is paid. If you want to broaden your horizons, Anthropic's Claude2 and Google's Bard are large-scale language models (LLMs) that have similar content creation methods but different aspects such as word count, web browsing, and data analysis abilities. may be suitable for each user. .

Twenty years ago, we would have needed a large number of apprentices. Ten years ago, training a custom model required consulting data scientists. Now it can be done in an afternoon with just you, Google Sheets, and ChatGPT.

Jonathan Soma, Knight Professor of the Practice of Data Journalism, Columbia University



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *