Mike Conover – Hello Dolly – Open Source Artificial Intelligence – A Threat to Big Commercial Models. Hashtag Trending Weekend Edition

So everyone already knows that I grew up in the 1960s. And one of my favorite songs by a band called The Who had “we won’t be fooled again”.

I thought about that when I was reading about Tim Berners-Lee thinking about how disappointed he was with the 30th anniversary of the World Wide Web. What should have been a free and open space became commercialized and largely run by large corporations.

Berners-Lee wasn’t the only one disappointed. Have we been fooled again?

Despite that disappointment, and many others over the years, some of us have never lost our sense of optimism or wonder.

It happened when one of our analysts called me and said: “You have to see this. Now, he said, I am playing.” He didn’t seem to have slept. “I’ve been playing with this all weekend. I can’t stop. It’s incredible.”

And it wasn’t World of Warcraft or some hot new game. It was…

… a really, really simple screen with one field that you can type in. You have entered the question in natural language.

Now, what makes that special? Well, I think we all know what makes it so special. It was ChatGPT.

We weren’t the only ones amazed. ChatGPT became the hottest app in the world, reaching a hundred million users in a fraction of the time of any other app. Even the mighty TikTok took more months than ChatGPT to reach that magical 100 million mark.

And it was available to you for free. And the name “open AI” was so catchy. It meant it was open source. Not true. “Fooled again?”

Nothing against Microsoft. But what appeared to be the democratization of AI was actually owned and controlled by a large corporation. And although others were coming into the field, they were also large companies. Google, Meta/Facebook etc will have their own AI.

I am not against commercialism. I am a capitalist. But I also love open source, that democratized vision of unbridled creativity and passion for doing something, not for the dollars, but for the intrinsic reward of doing something great.

Maybe it’s just a dream. After all, it takes billions of dollars to build and train AI. A big part of it is training. Maybe it’s time to face reality.

Maybe not. I read about another AI story. it was sponsored by a company but they promised to make the AI ​​model free and open source. Incredibly, they trained their model with a pool of only 5,000 people, in this case their employees who work as volunteers.

Could this be true? Could you really have a free, open-source artificial intelligence that a dedicated group could easily train?

I had to find out. Now I had another song going through my head and it wasn’t The Who, it was Louis Armstrong. She was singing Hello Dolly. You’ll get a link to it if you listen to the podcast.

My guest today is Mike Conover, and he’s from Databricks. And he’s been a big force behind their free open source AI they call Dolly.

UPDATE: After I did this interview with Mike, a new story started. I’ve covered it on HashTag Trending, our daily audio tech news podcast, and Tech News Day (YouTube) daily newscast. A leaked memo from Google made a shocking prediction that open source AI would surpass commercial AI. It was a very coherent argument.

Source link