0
0
Staying updated in the fast-paced AI industry can be challenging. Until an AI can do it for you, here’s a convenient summary of last week’s machine learning stories, including noteworthy research and experiments that we didn’t cover individually.
Google took center stage in the AI news cycle this week with a series of new products unveiled at its annual I/O developer conference. These offerings span from an AI code generator aimed at competing with GitHub’s Copilot to an AI music generator that transforms text prompts into short songs.
Many of these tools appear to be genuinely time-saving solutions, going beyond mere marketing hype. One standout is Project Tailwind, a note-taking app that utilizes AI to organize, summarize, and analyze files from a personal Google Docs folder. However, these advancements also highlight the current limitations and shortcomings of even the most advanced AI technologies.
Consider PaLM 2, Google’s latest large language model (LLM). PaLM 2 will empower Google’s updated Bard chat tool, serving as the company’s competitor to OpenAI’s ChatGPT, and serve as the foundational model for numerous new AI features. While PaLM 2 is capable of writing code, emails, and more, similar to other LLMs, it also exhibits toxic and biased responses to questions.
Similarly, Google’s music generator has its limitations. As I experienced firsthand, most of the songs created with MusicLM are passable at best and, at worst, reminiscent of a four-year-old playing with a digital audio workstation (DAW).
Much has been said about how AI could replace jobs, potentially accounting for the equivalent of 300 million full-time positions, according to a Goldman Sachs report. In a Harris survey, 40% of workers familiar with OpenAI’s AI-powered chatbot tool, ChatGPT, expressed concerns about it entirely replacing their jobs.
Google’s AI is not the ultimate solution. In fact, the company may be lagging behind in the AI race. However, it is an undeniable fact that Google employs some of the world’s leading AI researchers. If this is the best they can achieve, it underscores the fact that AI is far from a solved problem.
