Industry experts are constantly sharing their AI experiences and stories. We decided to cast the net wider and get a different perspective, in this episode we are joined by Chris Ehrlich, managing editor of Datamation. Ehrlich brings over 20 years of experience reporting on B2B technologies and chats with co-hosts, Evalueserve’s CTO Rigvinath Chevala, and VP of Marketing Erin Pearson about all things AI.
From the AI as a misnomer, biases, innovation demand, and more, the trio brings you an engaging conversation. Subscribe to the Decisions Now podcast today, you can find us on Spotify, Apple Podcasts, and Amazon Music among other platforms.
AI as a Misnomer
Often, the term AI is used interchangeably with different technologies. We asked Ehrlich his thoughts on the matter as someone who writes about AI and technology often.
“I think that it is a true misnomer. In that, we haven’t reached mainstream human-like intelligence,” he said. “And if that’s really the goal or the notion of the term, then I think generally speaking, in terms of mass market applications, we’re, as we know, far from that. And so, I think it’s overused.”
When companies talk about AI, often they mean Machine Learning (ML) instead.
“I think machine learning, by comparison, is a really apt description and accurate in that sense and really toward the end of automation,” Ehrlich added. “So, to me, it’s more straight ML and automation and AI is more of this sort of long-term goal. And I think it’s sort of overhyped essentially right now.”
Agreeing with Ehrlich, Chevala stated there is general AI, which is perceived as human-like intelligence, and narrow AI which is mostly ML. He added that he had recently seen a picture of a non-existent woman, generated completely by AI with intricate details down to freckles, the two-tone color of hair, and so on, which can be classified as the perception of general AI, which isn’t what typically applied to businesses.
“In pop culture, they use it very generally and it’s this really amazing human-like intelligence, it’s going to replace humans. It’s very sci-fi, kind of scary depending on how you want to look at it,” Pearson said.
The reality, however, when people say they have AI-based technology is often they are coding stuff, but they shouldn’t be using AI unless it’s truly an application of AI, she added.
With all the fluff in the verbiage around AI, the hosts asked Ehrlich his perspective on when in use when is AI legitimately AI versus when it’s just a marketing term?
“What I’m seeing is, typically startups are the most specialized and they are very application use case specific and technology specific,” Ehrlich said. “So, in terms of the marketplace, machine learning is what’s being used to deal with the abundance of data or the overabundance of data.”
In the marketplace, ML is being used to deal with the abundance of data, and that’s also where we see AI being used to solve big data-related challenges.
AI and ML
So why AI and ML being used interchangeably?
“AI and ML kind of go hand in hand in most material that you see out there. The way I see it is that machine learning is more of a mechanism. It’s the means by which you apply math to a certain set of problems. So, it’s in a way it’s an applied math problem and AI is the concept,” Chevala said. “It’s like two different things like, AI is just describing the fact that the outcome of using that mechanism, then mimics some sort of human intelligence, which is why it’s Artificial Intelligence. So, that’s the way I differentiate in my mind.”
Ehrlich said ML to him is what traditional coding is, what programmers or developers are doing at high levels when dealing with big data, from there AI is the next level, the graduation of that to when you can apply it in some form.
“You are in fact mimicking human intelligence or its human-like intelligence. And those use cases are so limited. And so if you look at chatbots as another kind of popular example and high-level example, yeah, there’s some of that happening and that’s sort of the beginning stages of that,” he added.
Biases
When you’re creating algorithms, they’re really only as good as the data that you can put into them, Pearson said.
“Whoever is working on AI or training the ML models, there’s going to be that implicit human bias, unless there are checks and balances in place at the hiring level, at the staffing level. And then within the technology or within the frameworks that they’re working with to say, “Okay. Well, there’s some kind of accounting for what you describe, where you have variants, or you have qualifiers, or you have factors that are a part of the process to where you can mitigate for that.” But I think, it’s going to always be imperfect, because science is imperfect and data is imperfect, and it’s always going to have that manipulative effect,” Chris added.
If a model is trained off synthetic data, the end results will be inaccurate.
“It’s hard, especially when we’re talking about consumer-facing AI and you’re dealing with this data that we were just talking about, the data is unpredictable. People are unpredictable. But if you train the models based on artificially generated data, then you’re completely off,” Chevala said.
While using synthetic data has its challenges as do historically entrenched databases that may include implicit biases.
“It’s a function of both the domain expertise that you build over time to avoid those biases and you have that critical massive data,” Chevala said as a potential solution to solving the challenges with bias in AI.
AI Advancement
Every time we blink, there seems to be another invention or technology. Pearson asked Ehrlich and Chevala where they are seeing AI advancement, in terms of application for B2C or B2B companies.
Fun applications of AI are often seen in robotic automation, like in retail beyond their prototype stages, but it’s a different question whether they are being embraced or widely adopted, Ehrlich said.
“But you have basically servers functioning in restaurants or bots traveling about in a restaurant, taking orders, returning the tray back, and so on. You can clean the interior of airplanes without basically a staff member essentially. They sort of set the machine in motion and it can clean the cabin. And so, these are sort of a marriage of automation and AI and robotics,” he added.
Whether there are advancements or not, it’s important to factor in whether the market is ready for a particular technology and if there’s a demand for it, the trio said mentioning Google Glass. The innovation was launched a few years ago, however, it didn’t grab the market, despite the product being impressive, but they are launching a new feature, powered by Natural Language Processing (NLP), that will translate language as fast as one speaking.
“That is a powerful use case. When you’re traveling, you have multilingual groups that you’re conversing with, that’s where they’re potentially going to hit product market fit,” Chevala said. “So, it’s a confluence of both technology, innovation and the desire to use that. I think that’s where most startups fail.”
Companies must factor in market demand to match their innovation.
Oftentimes with market demand at the consumer level, there needs to be almost an education, and awareness, there need to be these applications that really change people’s day-to-day lives, Ehrlich said.
As creators invent more products, the question of the hour remains – ‘is the market ready for this invention, yet?’
To listen to the full episode, subscribe to the Decisions Now podcast today!
