The Importance of the Human Element in AI

Thanks, Daniel and Jared. Glad to be back.

Oh, I totally appreciate that. I didn't know I got so many, reactions, which is great. One time, I was at a hockey game, and for some reason, I got on Instagram for, like, twenty five seconds, and there were, like, a million people. And I was like, I had no idea what was even happening. So I do appreciate the feedback.

One, I love my job, and two, this whole emerging field is something that I'm so interested in. So I'm I've really tried to, you know, become educated in it and talk to people.

And it's hard to keep up with all the different developments in AI and privacy, but I certainly try to see it holistically and how, you know, how I can bring some value to understanding it as somebody that works in a corporation, as a lawyer, you know, as just, you know, a friend, somebody who I can, you know, teach AI a little bit about and make it more relevant to people's use in their lives.

Sure. Well, I absolutely see AI as really making more efficient and really putting information together, tapping into sort of like a brain dominated analysis of information and, you know, selecting and analyzing and, you know, at some point, possibly, like, summarizing and early case assessment and things like that. So I think that there is a huge potential for AI in eDiscovery.

I do think, you know, we've gotta have the guardrails up in its use. There's gotta be ethical values surrounding its use, and it's not a free for all. I mean, I think that there has to be the trustworthiness, the transparency, the ethical use. I mean, I think that all has to be developed. And I think that I think that it is. I think that there's, I mean, I think there's a lot of attention being given to the governance and use of AI.

I'm seeing more, you know, more states and more country regulation of the ethical use of AI. And so I think that it exploded on our horizon in, like, so many areas of industry, of life, of medicine, of university, of scientist usage.

But I think now, you know, we kinda like we move forward, but we're also kinda like taking one step back to make sure that the governance and the integrity and the ethics are all there. And the trustworthiness is really the one of the most important concepts because, you know, if you're using a particular kind of AI, you wanna know for sure that it's, you know, proven to be trustworthy and defensible and transparent and explainable.

So I do think, you know, there there's a lot of that going on in terms of developing and demonstrating and fine tuning in the eDiscovery world and elsewhere in AI.

Well, I think it was so exciting, right, to have this technology and this is, you know, what we're talking about in legal week this year and etcetera.

And I think that, you know, there was a lot of experimentation and trials of AI and Gen AI, and people and entities were realizing, you know what? There there's a lot of risks with using this if it's not properly used.

So I think that the excitement and the emotion sort of was the first reaction to Gen AI and, wow, this is like you know, I was at a wedding and the father of the bride, he's like, I had chat CPT write my speech. You know? And we're like, you gotta be kidding me. But it was exciting.

Right? It was, like, right at the right time. And I read, you know, that scientists have used generative AI to either summarize or make introductions or whatever. And sometimes, like, it's not right.

Right? Because it depends what you know, it depends one, you know, what you feed into the foundational models.

And there's issues of, you know, entities have proprietary information that they don't want going into the ether to be used, you know, for open source, you know, Gen AI or whatever. And I think that there was then kinda like a look back period, and we realized, hold on a minute. Like, we have got to regulate this. We've gotta put guardrails around its use because, otherwise, it's you know, you're gonna get these hallucinations.

You're gonna get stuff that's not true. You're gonna get, you know, things that that are not accurate, and that's not the purpose of AI. Right? The purpose of AI is to provide for, you know, a more productive, truthful, efficient result or outcome, you know, in whatever it's being applied.

Like, if it's being applied for eDiscovery, you know, maybe identifying documents that are relevant and hits, you know, for a particular case. For medicine, scanning whatever billions of MRIs and PET scans and things like that. Okay. You want you want, you know, reliable okay.

This patient's MRI looks almost identical to this person who was diagnosed with this and got this treatment. So you don't you wanna you wanna operate, you know, in a in a truthful, accurate, regulated, ethical space.

And I think that we kinda jumped into it so quickly that, you know, many organizations, entities, governments realize, hold on. We gotta we gotta put the brakes on a little bit and really circumscribe the ethical use of AI, the guardrails, you know, what's appropriate, considering privacy, you know, considering all kinds of ethics, considering, you know, making sure there's not bias, making sure that the information that's put in all these models is, you know, good information. Right? Good documents, good records so that the AI, you know, works for the intended purpose and is not, you know, just a free for all kind of technology without any kind of oversight.

There's definitely you need human oversight, right, into creating the models, feeding the models, applying the models.

Obviously, the you know, the creation of the of these models, right, requires sensitivity to many issues, right, to human beings. Right? Wanna make sure there's no bias. There there's software tools now that have been created that you can, you know, layer up on your AI to make sure that there's no bias. But, for example, if AI is being used for any kind of employment decision, right, we wanna make sure that there's no discrimination, whether it's, you know, age or orientation or nationality, culture, whatever.

We wanna make sure that these tools are working properly. Right? And they're they're not inherently flawed. And I know, for example, there is use of facial recognition technology, and a lot of, you know, airports use it, airlines use it. I mean, it's, like, incredibly good technology.

And if you've been online at customs and, you know, international airports and things like that, you'll know that, like, you know, going into the screen and just, like, you know, they're recognizing your face and pulling up your data is a huge, like, efficiency. Right? You don't have to wait on line for two hours, you know, with your kids and the suitcases and everybody having to use the restroom where you can't go until after you've cleared that. But there were some facial recognition tools that had bias in it. Right? They don't recognize people of color that, like, some of the shading and everything just sort of is it generally identifies and is not particular.

And so we wanna make sure that, you know, AI doesn't have these, you know, these flawed elements, and we wanna make sure because otherwise, they're they're not gonna be beneficial to society. Right? For example, medical technology, you want diagnoses and prognoses and therapies to be available to everybody. Right?

You don't wanna just have AI be prioritized for, you know, white women or whatever it should be. This AI should work to the benefit of everybody in our world. Right? Everybody's created equal.

Everybody should have, you know, equal opportunities to get good care in industries as a client, whether you are, you know, a minority owned business or the biggest company in, you know, in Fortune five hundred.

AI should be available to your businesses to work in the best, you know, most ethical ways.

So I think that, you know, you ask how do ethics and human touch integrate into the use of AI.

These are just, like, some examples. Right? I mean and the other thing is, you know, human beings are not are not replaceable by machines, and machines are really helping us probably, you know, in many ways gonna be more efficient, live productive lives, help us at our jobs.

But, you know, you still need human beings. I had read an article last night about some scientists who said, okay, let me use, generational AI to draft something.

And there were completely disjointed sentences that were not reviewed.

And I'm sorry, but you know what? You still need to read what, you know, what you're asking for, see if it's appropriate, see if it's relevant.

And further, you know, AI is not a great use for, you know, everything we do in life. Right? Sometimes, you know, some some things are better done by the human mind. And so human beings won't be replaced, but AI can really, you know, pull information, data, databases across the world that, you know, we couldn't do otherwise, and then process a question to figure out, okay, what you know, get get an output and see, you know, what is coming as as the output and see if, you know, that can really help.

Yeah.

So it is an important topic. And, yes, like, I mean, human beings create their own problems. Also, you know, just the world does and different factors. We've created, you know, all the problems with the climate change and, you know, our environments and ocean coasts moving in more and different patterns and things like that. But human beings, right, I mean, we're, you know, we're trying to solve our problems, and we're trying to use AI to help solve our problems.

And I think that human beings are the ones and data scientists were the ones that, you know, created these language models to begin with and are creating these foundational models.

And so there's gotta be, as I said, like, these guardrails, right, on the models and the information that's put in the models and the selection of what the AI is being used for. I mean, at, at IBM, we have the emphasis on, you know, ethical use of AI and our, chief privacy officer and AI officer is on various governmental ethics boards and things like that because we have a very, aggressive stance that AI should be, you know, ethical, accountable, transparent, honest, sustainable, and, you know, be used with integrity. And so there's two factors when we're using it internally. Right? There's all these use case tests that you have to initially go through in order to ascertain that this is an appropriate use case for your business purpose or whatever.

Same thing, you know, when we're working with clients. Right? We we go through, you know, the whole litany of the client's concerns, intended use, information that's being fed, ethical use. And we have different sort of technology, like, packets that we provide to clients, right, that will embed it with, you know, anti bias, you know, all kinds of different software products that we layer up on the AI and make sure that the information is, you know, is ethical and the use is ethical. And so it's sort of like, you know, a laddering approach where, you know, you have, like, the foundational model on the bottom, and then you have different, you know, anti bias software. And, you know, you have other sort of guardrails on top and stacked up so that the product is, you know, gonna be the most, you know, transparent in terms of, like, having auditability, right, of the model, auditability of the data, the information, and the records that are being used. So it's always explainable.

There is data governance built into all the AI that we use internally as well as the AI that we use as client offerings.

And by data governance, I'm saying that there are and this is, you know, to work with the client, but determine what the retention periods are. Like, what we wanna make sure that, you know, there's no risk of misuse. And if there is some regulator asking about the AI, that there is the metadata available, the foundational model records available in an, you know, in a very organized, structured informational governance portal or whatever, however, you know, somebody maintains it. So I think that, you know, the design of the AI for IBM internally as well as for what we do for our clients is very multilayered, very much, you know, has these guardrails, has this transparency, accountability, integrity built into it, has this, you know, anti bias, you know, stacked on top of it.

And I think that this is not just happening. It it's by design.

It's by implementation, whether it's for internal use. We get so much education on the use of AI as IBMers. We were required last summer in, in August twenty twenty three to have about between twenty and thirty or forty hours of education on how to use a foundational model, how to create a foundational model, how to input data in the foundational model, how to query the foundational model, how to analyze the output of the foundational model. So we were all I mean, every single employee was required last year to understand how AI works. And I'm a lawyer. I mean, I'm not a, you know, a business person. I'm not a scientist.

I'm not a client executive or whatever, but I understand now, you know, the basics of how it works, and we have a plethora of offerings.

Thank you so much. I enjoyed it and, appreciate it. Have a great day.