The year 2023 has been all about artificial intelligence, with numerous developments in the field reported. The year has been all about AI advancement, from Google and Microsoft unveiling their own AI chatbots, Bard and Bing, to ChatGPT becoming even more powerful with GPT-4. Recently, Google released Gemini, which outperformed ChatGPT in several benchmark tests. The LLM now powers Bard and can be accessed by users all over the world.
According to reports, artificial intelligence has progressed to the point where it will no longer require human intervention to create new, smaller AI tools.
According to a Business Insider report, a team of scientists from MIT and the University of California collaborated with an AI tech company called Aizip to crack the code that allows big AI models to create little AI sidekicks on their own. Consider it similar to a big brother assisting a younger sibling, but with artificial intelligence.
Yan Sun, the CEO of Aizip, told Fox News that they’re using the mega AI models that power ChatGPT to create smaller versions automatically. They’re essentially on a quest to allow AI to evolve on its own. “Right now, we’re using larger models to build smaller models, similar to how a bigger brother helps [its smaller] brother improve.”
In addition, Yubei Chen, one of the team’s researchers, revealed additional information. He claimed that they had made a breakthrough by designing a fully automated pipeline for the first time. “It can design an AI model without human intervention in the process,” he went on to say.
Now, massive language models like ChatGPT drain the bank, costing a whopping USD 700,000 per day to run. However, these smaller AI models, known as TinyMLs, are less expensive and more portable.
TinyML can help with specific tasks. Consider how facial recognition, hearing aids, and even your home appliances could benefit from TinyML magic. Yan Sun elaborated on their mission, saying that they want to bring intelligence to everyday life, making it safer and smarter.