THE GOLD RUSH! TO GENERATIVE AI — GET YOUR ACT RIGHT
That Generative AI today is impacting, optimizing and accelerating organizations is beyond any doubt. However, adoption of generative AI is no cakewalk. It is full of obstacles. For now, let’s delve into two big obstacles.
Firstly, everyone is so focused on the end goal that they lose sight of the underpinning which is the data in the data quality. Many companies want to do AI. They want to have predictive models or use generative AI and creative ways, but the data that they want to use to source those systems is not organized or gathered properly. Having a good data strategy and and having good data quality to be used for these systems is a must for adopting generative AI.
The second obstacle is that of security and privacy. If you’re using and have a completely closed system that can contain all the data and if you’re not using any outside sources, the second obstacle is not as big an issue. But, when you’re taking advantage of some of these large language models like, chat GPT, the content, the prompts that you give to chat GPT and the content that you publish become part of its training model. This is where you need to be careful to keep your proprietary information proprietary and make sure that it doesn’t become part of the model that is trained to share with everyone else. So, these are the two big hurdles that people are wrestling with within adoption.
We talked a little bit about the risks. Now let’s talk a little bit more now about the ethical and legal responsibilities that we should be looking at when defining a comprehensive AI.
There are a number of ethical and legal responsibilities, for example, bias in generative, AI. It’s important that we should consider it and try to avoid bias in our use of generative AI.
When you think of how something is trained and the knowledge that it’s trained to absorb it is using public sources and that might not be right from a corporate point of view. Take, for example, the financial information about a company could be incorrect, untrue and, at times, hateful.
So, it is important that you be clear about what you are trying to do with your AI? How are you training it? What defines the truth?
Generative AI, if not used properly, may lead to a situation where people make take hallucinations for truth. Because, the information generated by generative AI is a piece of the puzzle that the machine figures out and leaving it. Something to put out as a truth. It isn’t the truth.
Generative AI is trained to generate. So when you present a question or a prompt to generate an AI, it’s going to give you an answer based on the information that it has. If it doesn’t have hundred percent information to fill in all the blanks, it will fill it in with something which is called a hallucination — which it takes a mixture of data and adds some additional data to make a complete response. But, that data is not based on any fact, anywhere.
Generative AI is trained to take what it has and give you an answer even if it has to fill in the blanks with some random content to make it sound good. We need to train it and design the system to parameters for use and operating.
How do we ensure that? I mean it still seems a little bit mysterious. There is really no magic in any of this. How do we train our systems to avoid these kinds of downsides, that seem to be almost out of our control.
In a digital transformation world, it all comes back to correlation data and quality data governance. Making sure that we are protecting information that should be kept private, that we’re not submitting it into sources so that it becomes data in the wild, is crucial.
Not only that, we should have a fair and complete idea of the tools that we are using with the content so that data privacy is maintained. We have a responsibility to at least be aware of how the tools work and whether they are going to make the data submitted to them public or not.
There are serious questions about who owns the content. What is generative AI actually able to do to re-purpose content? What is legal use, and acceptable use of intellectual property? I think those questions are going to take some time to settle out to figure out as more and more content becomes available.
In conclusion, if you’re thinking about AI and generative AI, consider data sources and your data strategies to make sure you don’t lose sight of the importance of the underlying data, and as a practical tool set.
Subscribe to DDIntel Here.
DDIntel captures the more notable pieces from our main site and our popular DDI Medium publication. Check us out for more insightful work from our community.
Register on AItoolverse (alpha) to get 50 DDINs
Join our network here: https://datadriveninvestor.com/collaborate
DDI Official Telegram Channel: https://t.me/+tafUp6ecEys4YjQ1
