Chapter 283: G.I.G.O.
A large U-shaped sectional sofa that faced a huge flat-screen TV dominated the family room of Jack’s mansion.
Jack sat in the center of the sofa with Rina on his lap, and Katie and Samantha on either side of him. The rest of his girls were also there, arrayed on either side of him.
They had gathered to watch the livestream of a financial news show where Heidy Oakland, his Radius 10K CEO, was scheduled to be interviewed.
The scene on the TV changed, and he pushed a button on the remote to unmute the show.
"In other news, a new contender for the AI crown is making waves, and we have the pleasure of speaking with their CEO this evening."
The speaker was an attractive woman in a business suit, and as she spoke, the screen split. Heidy appeared on the right half, also in a smart business suit.
"Welcome to the show, Mrs. Oakland," said the host of the show. "Your new company, Radius 10K, came out of stealth just two weeks ago and has already caused a lot of turmoil in the AI market. Some analysts are suggesting that you’ve already captured 20 percent of the market."
"Thank you for having me, Renee. And I think that number is a bit misleading. It’s closer to 30 percent now, but much of that shift has been away from paid AI services and towards using our new open-source model, Karl."
"Oh? Does that bother you? That you’re losing business to your own free model?"
"Not at all, Renee," said Heidy, smiling. "There is a huge demand for secure, private, local models, and we chose to fill that with free, open-source models and software. Our paid service focuses on scenarios that are still difficult for individuals and companies to solve with local models."
"Like your code review service and document review service?"
"That’s right, Renee. We’ve been focusing most of our effort on developing models tailored to specific tasks that humans find tedious but necessary."
"Like school homework?" asked the host.
"I would prefer that students not use AI to do their homework, but that horse escaped the stable a long time ago. We are looking at ways to mitigate that problem, but our solutions are still in development."
"Can you share any details?" asked the host hopefully.
"Sorry, not at this time," said Heidy, in a fake apologetic tone that oozed, ’sorry not sorry’.
Renee pouted, then asked, "Can you share any other future product plans?"
Heidy smiled and said, "Sorry, you’ll just have to wait like everyone else."
Renee sighed in fake disappointment, then asked, "Can you tell us anything about how your company managed to achieve such spectacular results? Your open-source model appears to be just as good as your competitor’s proprietary models, and your proprietary model, Radium, appears to be as much as 50 percent better."
"Sorry, Renee, I cannot reveal anything about our development and model training process. All I can say is that we use only public domain and licensed data."
"Interesting," said Renee.
Then she turned to face the screen and said, "Here to talk about AI training processes is the CEO of AI-n-stein, Hammond Saltzman."
The right side of the screen split again, with Heidy on the bottom and a frowning man on top.
"It’s a pleasure to be here, Renee," said Mr. Saltzman. "Mrs. Oakland, I’m surprised to see you running an AI company, given your history with the tech."
"I don’t have a problem with AI, Hammond, I just had a difference of opinion with Mr. Moor about how it should be used inside BuyMoor."
"Uh huh," said Hammond skeptically.
"So tell us, Mr. Saltzman, what are some ways a new company might train a new model quickly without incurring the costs a company such as your own has had to pay to reach your current success?"
"That’s an interesting question, Renee. One example is a problem that came up recently. We discovered that a competitor of ours, who shall remain nameless for now, was making use of our service to train their models."
"How does that work?" asked Renee.
"We hired many experts to solve problems so that our models could be trained on the solutions. Our nameless competitor was instead using our service to solve the problems, saving them the cost and time of hiring experts."
"Do you think other competitors might be doing the same thing?" asked Renee, leadingly.
"We are certain of it," declared Mr. Saltzman. "But identifying the offending accounts has been challenging."
"Do you have any thoughts, Mrs. Oakland?" asked Renee. "There are some who claim Radius 10K is just standing on the shoulders of giants."
"All of modern technology stands on the shoulders of giants, Renee," said Heidy. "There are numerous published papers that form the foundation for much of the current AI technology and we, like every other AI company, started with that knowledge as a foundation and worked to build on it."
"And what about using competitors to train your own models?" asked Renee.
"Garbage in, garbage out, Renee," said Heidy, smirking. "Unlike with humans, an AI student can only be as good as its teacher. So how would we ever exceed our competitors if we used them to train our models?"
Hammond glared at Heidy through the screen but remained silent.
Then Heidy added, "Unlike Hammond, we at Radius do not care if our competitors want to try training their models using our service."
Both Mrs. Saltzman and Renee looked surprised at Heidy’s statement.
"Are you not worried that your competitors will catch up to you?" asked Renee.
"Not at all, Renee. Not at all," said Heidy confidently. "We have a comfortable lead, and do not anticipate losing it."
"Bold words, Mrs. Oakland," said Renee. "Well, that’s all the time we have for this segment. Thank you for joining us, Mr. Saltzman, Mrs. Oakland."
"Thank you for having me," said Heidy.
"It was a pleasure as always, Renee," said Mr. Saltzman.
Heidy and a still-frowning Mr. Saltzman disappeared from the screen, and Jack muted the sound.
"I think that went well," said Miranda.
"Yes," said Nora. "I especially liked the ’garbage in, garbage out’ comment. I think that’s when that vein on Mr. Saltzman’s forehead started throbbing."
"How long do you think it will take them to catch up?" asked Samantha.
"I guess it depends on how quickly they manage to bribe someone at BallSoft or Yoctoly," said Madison. "Not that it will matter. As soon as they catch up, we’ll release a new model."