Learning Library

← Back to Library

Become Your Own AI Firestarter

Key Points

  • Fire transformed early humanity by providing light, heat, and new technologies, and Dario Gil likens generative AI to a modern, shareable “fire” that can similarly unlock societal progress.
  • Most organizations already use “traditional AI” embedded in off‑the‑shelf tools for narrow, task‑specific functions that require manually labeled data for each use case.
  • Generative AI, built on foundation models trained via self‑supervised learning on massive text corpora, can perform a wide variety of tasks from a single model, eliminating the need to develop separate models for each application.
  • AI can be consumed in three primary ways—embedded in vendor software, accessed as a standalone service, or customized and built upon directly—allowing businesses to move from being mere AI users to creators of AI‑driven value.

Full Transcript

# Become Your Own AI Firestarter **Source:** [https://www.youtube.com/watch?v=fBR42OxjWaM](https://www.youtube.com/watch?v=fBR42OxjWaM) **Duration:** 00:14:21 ## Summary - Fire transformed early humanity by providing light, heat, and new technologies, and Dario Gil likens generative AI to a modern, shareable “fire” that can similarly unlock societal progress. - Most organizations already use “traditional AI” embedded in off‑the‑shelf tools for narrow, task‑specific functions that require manually labeled data for each use case. - Generative AI, built on foundation models trained via self‑supervised learning on massive text corpora, can perform a wide variety of tasks from a single model, eliminating the need to develop separate models for each application. - AI can be consumed in three primary ways—embedded in vendor software, accessed as a standalone service, or customized and built upon directly—allowing businesses to move from being mere AI users to creators of AI‑driven value. ## Sections - [00:00:00](https://www.youtube.com/watch?v=fBR42OxjWaM&t=0s) **Igniting Your Own AI Firestarter** - Dario Gil compares fire’s historic impact to generative AI, urging organizations to transition from passive AI users to proactive creators who control and generate their own AI‑driven value. - [00:03:06](https://www.youtube.com/watch?v=fBR42OxjWaM&t=186s) **Three Modes of AI Consumption** - The speaker explains that a single foundation model can serve many tasks and outlines three ways to consume AI—embedded in off‑the‑shelf software, accessed via API calls, and (implied) custom deployment—showing how each impacts competitive differentiation. - [00:06:14](https://www.youtube.com/watch?v=fBR42OxjWaM&t=374s) **Adopting a Platform AI Model** - The speaker argues that businesses should use a platform approach—leveraging pre‑trained foundation models, customization tools, and integrated processes—to build and own unique AI solutions while still incorporating third‑party AI where appropriate. - [00:09:21](https://www.youtube.com/watch?v=fBR42OxjWaM&t=561s) **Multi‑Model AI Governance Imperative** - The speaker stresses tight data control and robust AI governance, advocating a platform strategy that mixes proprietary and open‑source, fine‑tuned models to deliver safe, sustainable business outcomes. - [00:12:35](https://www.youtube.com/watch?v=fBR42OxjWaM&t=755s) **Empowering Businesses with Generative AI** - The speaker urges business leaders to embrace generative AI—partnering when necessary—to become “AI value creators” that drive long‑term success, solve complex societal challenges, and unlock new, responsible solutions. ## Full Transcript
0:00When oxygen, heat and fuel combine, we get fire. 0:04It's basic. 0:05It's primal. 0:07And is the key that unlocked human progress. 0:10Fire provided light and heat and protection. With fire 0:14our ancestors could move to new climates and eat new foods. 0:19Pottery, metallurgy, chemistry, rapid transportation 0:23and many other technologies all start with fire. 0:28But imagine if fire had been proprietary. 0:31What if that knowledge hadn't been shared? 0:34What if there had been just a few keepers of the fire? 0:38What would we be? 0:57My name is Dario Gil 0:58and I'm senior vice president at IBM and a director of IBM Research. 1:04We are in a landmark moment now with generative AI 1:07that will shape our society for generations to come. 1:11In this episode, I'm going to show you why you need to be your own 1:16AI firestarter, how to take control of your AI destiny, 1:20to go from being just an AI user to become an AI value creator. 1:25Chances are you're already an AI user, even if you don't have a formal 1:31AI initiative or an AI team in place, it's already in use across your business. 1:37It might be baked into your off the shelf applications. 1:40It may show up in chat bots, in H.R. 1:42self-service portals, and transcription services. 1:46These are what we might call traditional AI. 1:50The focus is on executing discrete tasks, and usually each instance is trained 1:54individually on its own data that you have to compile and label. 1:59Now, what makes Generative AI different is that it is enormously flexible 2:03and not limited to narrow tasks. 2:06Think of it as instead of filling specific blanks, 2:09it can write the whole document from scratch to the point of being able 2:12to create or generate something entirely new. 2:16What makes this possible, 2:18are foundation models. 2:20Foundation models aren’t trained the same way as traditional AI. 2:24They're trained using self-supervised learning. 2:28You don't have to manually annotate a massive amount of data. 2:32You tell them to go read enormous amounts of text, and they do. 2:36And you end up with a large but versatile model 2:39with more humanlike language capabilities. 2:42Algorithms use mathematical models to represent 2:45the relationships between the words they ingest. 2:49If you give the model a few words in a prompt, 2:52you can mathematically predict the likelihood of words in the response. 2:57Instead of needing to build one A.I. 2:58model for each specific task, you can train one model 3:02and adapt it to many varied downstream tasks. 3:07So we went from one task, one model to one model, 3:11many tasks. 3:12Your chat bot and your H.R. self-service 3:15can be built on the same model as the new app 3:18that will write your marketing emails and summarize legal documents. 3:23So that's the first critical point. 3:25Ideally, a model is in the final form of AI. 3:29It's just a foundation you build on. 3:32How you use it is up to you. 3:38When it comes to using AI, 3:40there are basically three modes of consumption. 3:43The first is embedded AI, 3:45which I already mentioned, is baked in to off the shelf software. 3:50The software vendor creates the AI 3:53and you put it to use in your business. 3:56Whether it's a writing a system that can help you strike the right 3:59tone in your email or image editing software 4:02that can automatically process your images, 4:05you get access to some great functionality 4:07that can make you more productive, which we always want. 4:12But the caveat is that what you can buy 4:14so can everybody else. Those capabilities 4:17and productivity opportunities don't become differentiators. 4:21They set a new higher baseline for everyone. 4:25The second mode of consumption is through API calls. 4:29As you develop custom applications for your business, 4:33they can call out to another company’s AI service 4:36using that company's models and processes and then return the results. 4:41This is also a valuable way of consuming AI, 4:45depending on how cleverly you use the APIs and the diversity 4:48of AI service providers you use. 4:51You can start to differentiate how you put AI to work relative to your competitors. 4:57But there are caveats here too. 4:59The first is that, like with software, 5:01the models and services you tap into are available to everyone. 5:06The second is that when that API call goes out, 5:10it's connecting to what looks like a black box. 5:13You don't necessarily know what's happening on the other end 5:16or the provenance and governance of the data use, 5:19which can make people nervous because your business 5:22is still accountable for the final outcomes. 5:25And a second word of caution when using someone else's 5:29AI, has to do with the creation and accrual of value over the long term. 5:34In the past, we've seen a lot of value extractive business models. 5:39Another company would offer you a service like this API call 5:42and you get value from that. 5:44But the other company is also extracting value 5:47from your usage and from your data accumulating more and more. 5:52How much faster is their value growing than yours? 5:54Well, you can see it in the stock prices over time. 5:58There is an imbalance in the relationship, 6:01that can have long term consequences both for your specific business 6:06and for the overall economy and progress of technology. 6:11This goes back to the metaphor I used earlier. 6:14Philosophically, do we as a society really want just a few keepers of the fire 6:19upon which we are all dependent? 6:22Is that what's best for your individual business, for your shareholders? 6:27The third model of A.I. 6:29consumption is the platform model. 6:31This is the most comprehensive. 6:34This is how you become your own AI firestarter. 6:38And no, it doesn't mean doing it alone and reinventing the AI from scratch 6:42and spending years and millions of dollars to build your own models. 6:48With a platform you have all the elements and ingredients 6:51in place to build your own AI solutions. 6:54You have foundation models. 6:56You have tools to improve and customize models, 6:59and you have processes to build your own tailored AI solutions. 7:04And importantly, you create 7:06and accrue value that is unique to your business. 7:10Ultimately, I believe that most businesses 7:13should end up with a mix of all three. 7:15You'll use third party software with AI embedded. 7:19Sometimes you'll be totally appropriate to use someone else's AI, 7:23but to fully realize the value of AI and to differentiate yourself from competitors, 7:28you’ll want a platform approach to create and own your AI. 7:37Let me just go a bit deeper 7:39on AI value creation, starting with foundation models. 7:43Foundation models are large scale deep neural networks trained with 7:47lots of unlabeled data and subsequently adapted to many downstream tasks. 7:53It may be a broad general model, or it may be a narrower, deeper model, 7:57but the key is that it is pre-trained with the expectation 8:02that you can further enhance it with your own proprietary data. 8:06Like when a new employee joins your business, 8:08they come in with some general skills as a foundation and the ability to learn. 8:14The more they learn about your business, the more they add 8:17institutional knowledge and expertise, the more value they deliver. 8:22Well, the same is basically true of our foundation model. 8:25You use your AI platform to tune it with your specific business data, 8:30your proprietary knowledge and expertise, 8:32and it becomes more expert and more valuable for your business over time. 8:37And because you're in control of the platform and the processes 8:40and the data, you accrue ever larger amount of value over time. 8:45With some of the consumer AI on the market, 8:48we've already seen some of what happens when you surrender that control. 8:53You can get bad data that leads to bad outcomes. 8:55You can get hallucinations, basically an 8:58AI generating very confident and very incorrect answers. 9:02You can get into some trouble for inadvertently 9:05using someone else's rights managed content. 9:09We've even seen proprietary or sensitive data 9:11being inadvertently leaked back into the public spaces. 9:15That's why you need to know how your model is built 9:18and what data was included. 9:21And it's why tight control of your sensitive data should be prioritized. 9:26Strong AI governance is absolutely critical. 9:31Yes, now it is the time to jump into AI, 9:35but please look before you leap in and ensure 9:38that you're investing in a smart, safe, and sustainable approach 9:42where your business and your clients and customers 9:45are the primary beneficiaries. 9:51There is, I think, a myth about AI right now, 9:55a basic misunderstanding. 9:57For the general public generative AI has seemingly come out of nowhere. 10:02A lot of people think that there is 10:04a handful of consumer oriented AI experiences out there 10:08and that one model is going to win, or just a few leaders. 10:13I don't think that that's how it's going to play out. 10:15The future of AI is not about one model, 10:18it's multi-model. 10:20Your business will be using multiple fine tuned models 10:23to achieve the best results when applied to specific use cases. 10:27That's why that platform approach is so important. 10:30And realistically, the future of AI is not only proprietary. 10:36It will also be powered by open science and open source. 10:40Proprietary models will play a part. 10:43But so much of what is going to happen in the future will not happen behind closed doors. 10:49It will play out in plain view with full transparency and accountability 10:54that open source provides. 10:57The energy in the open source community is phenomenal. 11:01They are distributed projects, university projects, corporate efforts, all driving 11:06innovation, producing foundation models that you can tune and deploy 11:09for your use cases. Hugging Face is like GitHub for foundation models. 11:14And there are over 325,000 open 11:17source models available and thousands more being added. 11:21And this is exactly how it should be for the good of society in the long term. 11:26We don't want just one or a few winners, a few companies 11:30that can define what AI is and dictate how it's used. 11:34But that's not going to happen. 11:36What we're going to see is the democratization of AI. 11:44Okay, 11:45so I've talked about a lot of different topics, and I basically told you 11:49that you have a lot of work to do to be great at AI, 11:53and maybe that feels overwhelming, but really that's not my intention. 11:58My goal is to take down barriers to participation. 12:02Not put them up. 12:03There is a sense of urgency and fear about waiting 12:06too long and missing the moment. 12:09That's okay. 12:10I assure you that everybody is in the same situation, feeling the same emotions. 12:16But I promise that if instead of rushing into fast 12:20and easy options to check the box and say, yes, I put AI in the business, 12:25if instead you are thoughtful and deliberate and strategic 12:28about an AI platform and data management, and AI governance, 12:33if you become an AI value creator. 12:35You're going to be in a position to succeed over the long term 12:39and you won't have to start over every time the wind changes direction. 12:44You may not have all the products and processes you need today. 12:48You may not have all the skills in place. 12:51That's fine. 12:53If you're not ready to start alone, 12:55find a partner that you can trust and get to work 12:59so that you can take advantage of the extraordinary 13:02opportunities that generative AI will unlock for your business. 13:07Personally, I'm 13:08very excited about this chapter in technology. 13:12We, all of us together are going to use the AI to reshape 13:17not just our digital world, but our physical world. 13:20We are going to use it to help tackle some of our toughest 13:23social, medical, environmental problems, and more. 13:27We'll do it through science, but also by empowering businesses 13:31like yours and mine to do more, faster and more responsibly. 13:36Whatever things you do, you're going to get 13:39a power full new tool to help you do them better. 13:42And its the AI value creators, 13:44that will be the ones who make the biggest impact. 13:48It is the value creators who will take the amazing foundational technology 13:53that is generative AI and use it to build entirely new solutions. 13:57That's why it is our goal to make AI accessible to everyone. 14:03To put it in your hands 14:05and I can't 14:06wait to see what you will do with it. 14:09Thank you for watching this episode of AI Academy. 14:13I hope you will join us again as our IBM experts take you on deeper dives 14:18into everything that a business leader needs to know about AI.