by Mary Wells Posted in Business | December 06, 2024 3 min read Artificial Intelligence promises to transform lives and business as we know it.But what does that future look like? The AI Forecast: Data and AI in the Cloud Era, sponsored by Cloudera, aims to take an objective look at the impact of AI on business, industry, and the world at large. Hosted weekly by Paul Muller, The AI Forecast speaks to experts in the space to understand the ins and outs of AI in the enterprise, the kinds of data architectures and infrastructures that support it, the guardrails that should be put in place, and the success stories to emulate…or cautionary tales to learn from.AI is only as successful as the data behind it.
To explore what the next era of data looks like in this AI boom, R “Ray” Wang, principal analyst, founder, and chairman of Constellation Research, joined us to kick off this new podcast and discuss. Here are some key takeaways from Ray in that conversation.LLM precision is good, not great, right now Paul: I wanted to chat about this notion of precision data with you.And specifically, I was reading one of your blog posts recently that talked about the dark ages of data.
Walk us through where we are with precision data today and how this relates to the dark ages of data.Ray: We’re at a point where people get excited about 85% accuracy in their LLMs.85% accuracy for customer experience means that number isn’t bad.
What does that look like? You may get a telemarketing call and it gets routed to the wrong person.Or you might get an extra fry by accident at the checkout.These are all minor. But 85% accuracy in the supply chain means you have no manufacturing operations.
85% accuracy in finance can put you in jail.Therefore, the next 10%, which are small language models, are going to come into play.And the value of the 10% is as much as the 85% and as much as the next 5% to get to 95%.
To get to a full 100%, that last 5% is even more valuable.That’s context, that’s location.It could be metadata that you weren’t capturing before.
That’s anything from perspiration to heart rate – it’s all being captured.The final hurdle to LLM precision, available data Ray: But to get to a level of precision that your stakeholders are going to trust, there’s not enough data.Most of the publicly available information on the internet has already been scrapped.
There’s nothing new.People aren’t putting stuff out there anymore because they’re afraid.We went from not having enough data, to having all the data we know, to after 2022 not being sure what happened because people started hoarding data. We are going to enter the dark ages of data and the internet because nothing of value is going to be available publicly.
Value chains emerge in the midst of Dark Ages Ray: Given the dark ages of data and the internet, all the new information and insights are going to be worth something.You’re going to value your company not just by the revenues, but also by the business graph and the data that’s behind it.Companies will partner, but not with each other in terms of competitors.
A big retailer might partner with the manufacturer and a distributor to share information on demand or intervention on pricing elasticity or about available supply.That kind of information is going to become very valuable, and people are going to bid and build markets against that. Data collectives are going to merge over time, and industry value chains will consolidate and share information.It’s not direct competitors.
Retail manufacturing distribution is a natural value chain.These natural value chains are going to start learning how to share data and use different mechanisms to do that.+++ Don’t forget to tune in to Spotify or Apple Podcasts to listen to future episodes of The AI Forecast: Data and AI in the Cloud Era.
Mary Wells
Chief Marketing Officer
More by this author
Editor's Choice
Business
Acquisition of Verta’s Operational AI Platform Will Transform Cloudera’s AI Vision to Reality
Business
Bringing Financial Services Business Use Cases to Life: Leveraging Data Analytics, ML/AI, and Gen AI
