Searching for Meaning in the Age of AI
Summary
- AI meaning crisis exploration: Bryan McCann explores how AI is reshaping work, creativity and human identity while simultaneously offering inspiration and solutions to existential questions.
- Philosophical AI integration: This video discusses using AI responses as starting points rather than final answers, building teams like neural networks optimized for learning and continuous adaptation.
- Humanity's new story: McCann argues that humans must author humanity's new narrative in the age of AI, using AI as inspiration to become something new rather than fear replacement.
Searching for Meaning in the Age of AI

Talk by Bryan McCann, CoFounder and CTO, You.com
Bryan McCann, You.com’s co-founder and CTO, shares his journey from studying philosophy and meaning to the Stanford Computer Science Department working on groundbreaking AI research alongside Richard Socher. Right now, AI is reshaping everything we hold dear — our jobs, creativity, and identities. It’s also our greatest source of inspiration. The Age of AI is simultaneously a Renaissance, Enlightenment, Industrial Revolution and likely source of humanity’s greatest existential crisis. To surmount this, Bryan will discuss how he uses AI responses as new starting points rather than answers, building teams like neural networks optimized for learning and how the answer to our meaning crisis may be for humans to be more like AI. Exploring AI’s impact on politics, economics, healthcare, education and culture, Bryan asserts that we must all take part in authoring humanity’s new story — AI can inspire us to become something new, rather than merely replace what we are now.
Databricks Named a Leader in the 2025 Gartner® Magic Quadrant™ for Data Science and Machine Learning Platforms: https://www.databricks.com/blog/databricks-named-leader-2025-gartner-magic-quadrant-data-science-and-machine-learning
Build and deploy quality AI agent systems: https://www.databricks.com/product/artificial-intelligence
See all the product announcements from Data + AI Summit: https://www.databricks.com/events/dataaisummit-2025-announcements
FAQs
What is the meaning crisis that AI creates?
AI's capability to automate complex work threatens what humans identify with as valuable and meaningful, raising questions about purpose, creativity and human identity in an AI-capable world.
How should humans respond to AI capabilities?
Rather than fearing replacement, humans should treat AI responses as starting points for further thinking, build collaborative teams optimized for learning and use AI as inspiration to evolve.
What role should humans play in the age of AI?
Humans must actively author humanity's new story by engaging with how AI transforms politics, economics, healthcare, education and culture, becoming co-creators of the future rather than passive observers.
How can AI inspire rather than replace human capabilities?
By viewing AI as a tool for augmentation and learning, humans can develop new skills, pursue deeper meaning and evolve alongside AI technology rather than being displaced by it.
Full transcript
This is a little bit more of a uh philosophical conceptual discussion. I'm saying I'm saying up back there. Uh I'm going to breeze through the forward-looking statement here and get right into the main topic, the idea of searching for meaning in the age of AI. Uh the main question that I've been hearing for the last year and a half, two years from a lot of people as I've been talking to a lot of different companies and individuals is, well, how good is AI really going to be?
Is there anything AI can't do better than humans? And you don't maybe know anything about me yet, but I'll try to show you that my answer might be a little bit credible when I say probably not. uh kind of kind of sort of you know at least in theory uh AI beat us at our own games many of you are probably familiar with this it mastered our language to some extent now it's even taking our jobs and that's why this question started coming up a lot more um because it's not just the dull jobs right it's not just the jobs that people don't want uh it has the potential to some extent to annex all we identify with being valuable uh human beings and leaving living meaningful lives to some extent.
Um talk's going to stay this philosophical so just giving you a warning uh this not going to be super technical. So what I want to talk about is what do we do and for everybody in this conversation like what can we do what can we do with others uh and I'm going to draw from themes from my own search for meaning which started in philosophy but quickly became AI research so my background was as an AI researcher before I started my current company you. com uh and when I was studying philosophy I was really interested in what the heck this whole meaning thing is there's meaning of words individually and sentences and there's a lot of like linguistics and things like that but we can have a meaningful conversation we can have a meaningful memory meaning is this vague term up there with consciousness and other things like that I really wanted to try to refine and understand better uh and it turned out most of that gets translated into philosophy of language and analytic philosophy and eventually I thought that was kind of a dead And so I was studying computer science on the side for fun and wanted to see if we could study meaning with machines and even maybe make machines that could make meaning.
The spoiler is that I never really found meaning anywhere in those machines. But there were some key themes that I think are really important for me now and over the last 10 or 15 years working in AI that have helped me predict the future and then understand how to change in response to all the things that are happening. So the first was back when I was getting exposed to machine learning and deep learning. It was right when we were making this big transition from manually based like feature engineering and coding up very specific types of algorithms to try to help the machines know what to do and what was important and instead this is one of the first themes stepping back getting out of the way and just setting up more simple generic learning algorithms to let the machines figure it all out.
So the transition from machine machine learning to deep learning was in some sense humans getting out of the way focusing a little bit on the higher level objectives and not so much worrying about a task specific or really any sort of like narrow thing. My first real research paper was on something called contextualized word vectors. And the main idea here was that meaning is heavily heavily dependent on context. At the time, you know, we weren't using a lot of this context.
This is long before large language models, before language models were really even a thing. They were more a toy that were hard to publish on at this time. And ever since then, we've been introducing ways to get more and more context into our systems for AI. Very broad principle, I understand, but many people are probably familiar with context windows and extending those and now we have million token context windows or even more.
Really, so much of what we've done and the ways we keep progressing are by incorporating more and more context. My second big paper for my research days was on creating a single architecture that could do all of the different tasks in natural language processing instead of building an architecture specifically for one task. So, it's similar to that first thing I was talking about. Um, trying to get out of the way a little bit and realizing that tasks at least in the context of NLP and now to some extent tasks in the context of building agents and training language models, they're mostly made up in some sense.
Like our conceptual biases frame what these tasks are, but in actually making more capable AI systems, it's been better to not really think about that. Fun fact, maybe some of you disagree with this idea. At the time, it was heavily contentious. Uh could never get this published. Uh tried many times, worked on it for about a year and a half, submitted to many conferences. There was always at least one reviewer that said, "No, it's a terrible idea."
Uh there's no such thing as like general question answering systems or general systems that can do it all. But there was one group that really really liked it. And that was OpenAI. So they cite this work as uh one of the first works demonstrating that language itself could be our interface to AI systems to neural networks and that could be how we interact with them instead of designing them for specific tasks back out of that and use language as the main medium.
So this is the GPD2 paper from openAI you know we're on GPD 3 4. 5 etc. This is back in the GP2 days. After that, I went and applied all these language models to biology, uh, dialogue, chat, basically every area that I could get any data on, any context on and started seeing what would happen. Now, I want to summarize a little bit. So the first theme was that we should bet on simple scalable learning over any sort of dogmatic conceptual biases even if they seem obvious to us.
Right? When people were telling me it was a terrible idea to make large unified NLP systems, they were coming from a place of engineering principles really. It was going to be super hard to create these systems to deploy these systems. They'd be slow. It wouldn't make sense for users for all sorts of reasons. first principles thinking well depends a lot on your principles and it kind of betrayed what happened in the late the years that hap came after with openi and the rise of large language models so I think we should bet on learning more than our own biases we should incorporate as much context as possible into any sort of decision-m all these will come back don't worry and AI is going to change everything so this is the mode I've been in for the last seven eight nine years everything.
Biology as we were showing, uh, conversation as we've all seen now, economics, which is maybe not changing quite so much yet, but a lot of potential, politics, society, quite literally, like what it might mean to be human. I'm expecting these things to push us further and further to try to understand what, if anything, is unique about being human. And we have an opportunity to reinvent all this stuff. So, it's a little scary for some, but I really see it as an opportunity to reshape what all of these things could look like.
Where are we at? Where are we now, though? Uh, well, there have been three waves so far. We focused on just being able to generate answers. This was the era of co-pilots predominantly and now we're in this age of trying to make agents themselves giving the models more agency allowing them to do and decide things without us interfering which is a great step in that direction of reducing our own conceptual biases interfering with what what they what they might be able to do.
This is giving them more task autonomy, but we really want to get to complete autonomy, full work products, completely automating a lot of workflows. And what's missing there is still to some extent a lot of context. So when we build these enterprise search and agentic systems at. com and many other places, however you might do it, if you're building these compound AI systems based on those themes I mentioned before, there are two things we should notice. fun, a lot of this is just bringing in more and more context in different forms, different data sets, like what whatever we can get our hands on.
And the other, which is in the back of my mind, is this is starting to look more and more complex and a little bit less simple and scalable. So, the pendulum tends to swing back and forth on these things. And I'm expecting at some point a lot of these compound systems that we're building now that are really getting us really far are going to get replaced by something simpler, easier to learn. That's okay. This is just how it works.
We go back and forth until we kind of gather enough context to realize what might be going wrong and get enough intuition as to how we might do this all in a better, simpler, more efficient way. But even within that, as you're building these, we try to design everything as much as possible for learning and get out of their way. That's how at least at. com, we can still build these agents that are really, really effective.
Even compared to like giant efforts like OpenAI, we can do better in different areas on things like instruction following, comprehensiveness, completeness. The way we're doing it is by bringing in more context, focusing on these different learning uh strategies and you'll see this leaprogging and like jocking for position across OpenI anthropic and everybody else, right? Because everybody is on to this theme that you can keep doing this to make the models better.
Even with data bricks, which I must I must mention, right? We're at the data bricks conference. We've been data bricks customers for a long time. Uh we've built essentially all of you. com on data bicks. Our partnership with them that we just recently announced involves allowing all of data bricks customers to access our APIs for general web search for news for image search coming up as well. This is a way to give everybody's LMS and agents more context so they can stay up to date with the rest of the world since they stopped once they stop training.
They don't know what's going on. as soon as one more day happens, they might not know what's going on in the world. And more and more people now, they have their basic rag setups um more or less figured out, they're moving to this next step of realizing the models are up are out of date typically on their own, and they need these kind of search APIs to be able to make reliable and trustworthy decisions. So, if we're going to get to that full autonomy state, we really need to have them be able to access proprietary data within a company, but also know what's going on in the world.
And then in addition to that, we're collaborating with data bricks on like a few other efforts that you'll see coming out over the next couple quarters on our agents, additional AI infrastructure tooling. All of this is different ways to give these models more context. The other thing that we found really important in the kind of vein of these talks as well is working with people. So training people to understand how to use these systems has been crucial to get adoption.
It's very hard for a lot of people to make this conceptual shift that in my mind has been happening for 1015 years all at once and then get the maximum amount of value out of these agents and LLMs. If you just spend some time with them, maybe one dispersing some fear that they have because a lot of people are still afraid of these things that they're just going to replace them and take their jobs away. Then you get an empowerment phase instead.
You get champions. You get people thinking about how their jobs could be different. Thinking about how they could be moving more and more towards their ideal job or role or function in an organization, maybe even how their organization could change. So spending time with people is really really crucial for us as we've seen in making sure all of our systems can be successful. And then what we're seeing from that is that companies then build our tools in very transformative ways for themselves.
The DPA is the German press agency and they've used all of our tooling to essentially create a replacement tool for their writers. And now their writers just use a version of you. com on 10 20 million articles that they've written over the last several decades to get drafts. And that's like a very different and risky proposition for them. But as they're embracing this kind of transformation, they can turn into a different kind of company.
WBV is similar, also a media company, but instead of just building internal tooling, they've actually turned they've actually added all of this as a new product for their customers and they're bringing in more revenue. So, this is like a way for them to potentially over time become a completely different kind of company. And this is really just a set of point examples along that dimension of trying to transform as much as possible.
I really think if we're bold enough, all these jobs and all these companies as we know them will either be automated or transform into something completely different over the next 10 years. A lot of them might go away, but most of them have an opportunity to transform. I also think that most systems and institutions can change for the better as well. Again, if we take advantage of the opportunity and we're bold enough, and as I was saying with that theme earlier, we could reimagine a lot of political and economic systems in a quote unquote jobless world where maybe we still have very important roles, the things that are valuable to us, things that give us meaning in life.
Every once in a while when I give versions of this talk, people come up to me after and want to talk about, you know, what is left, especially if they identify with their job or their work or what they're doing a lot. But the answers are always very similar. It's community, it's connection, it's expression, it's creating. There are things that are experiential values that humans seem to just be drawn to and find meaningful even if we lived in a world that was completely different.
Maybe this is giving us some hints of where we're going to be different from AI. No matter how AI how good AI gets at doing things, there's something here. This is not a fully fleshed out idea, but it's something that I'm enjoying sparking conversations about because I think it's going to become easier and easier to create a world according to your ideals and vision and imagination. All the execution can become automated to a great extent.
So, I would encourage everyone to spend at least a decent chunk of their time now envisioning that future and trying to get a sense of where you want the world to go because it's become more and more uh uh a discussion, a negotiation amongst all individuals about like what that should really look like. So, how do we prepare for that? Well, I think the answer is to be more like AI, which I I don't know, maybe that's uh uh sounds paradoxical, but had this metaphor of comparing individuals and cities to try to flesh out the idea of justice.
Uh I think neural networks, artificial neural networks and human neural networks, we already have this metaphor in place. So if we look at what has been really successful for these artificial neural networks, maybe we can learn from that ourselves and we can be more like them. And what were those things? Well, it was a shift away from those task specific mindsets. So in our jobs, in our roles, in the things that we're doing, shift away from that towards more dynamic adaptability, focus on continual learning. what it's going to mean to be good at your job or whatever you're doing as a human is going to be much more based on that than it is any particular thing you're good at because if you're good at it probably AI will be better at it soon.
Incorporating more context into what it means to be human might be a very abstract idea, but I would really encourage everyone to think about all of the ways in which we might feel human or implicitly or unconsciously think about being human that we don't always verbalize. There's a lot of like consciousness or like how we experience things that we don't have good language for yet. And just because we don't have good language for it yet, we don't really have good ways to talk about humans being different than AI in any way.
But actually articulating that language and creating that language is something that large language models aren't particularly good at. But we can do that for ourselves. And maybe if all else fails, I think perhaps what will be left for us is just like the search for meaning itself. It seems to be something that uh humans need on some level or at least want. There's like an aspect of narrativization uh story uh the search for whatever meaning is and trying to understand it and create it is guiding us in some direction.
I don't know where but if we keep following this I imagine we're going to continue to refine what it means to be human. we'll have this opportunity to create all these different institutions, a completely different world. And I hope, you know, as we finally unlock the potential of AI, AI can help us unlock the potential of humanity. And that's all I got for you. Thank you.
Learn more about the Databricks Data and AI platform.
The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.