Why (Senior) Engineers Struggle to Build AI Agents

Talk transcript

1. 0:00

   [upbeat music] Okay, cool.

2. 0:15

   Awesome. Hi. Hi, everyone. Uh, my name is Philipp. I work at DeepMind, uh, everything related to agents on Gemini or Gemini API. So if you have some questions afterwards, some concerns, some bugs, some issue, please let me know.

3. 0:28

   Uh, we're going to talk today ten minutes about why engineers struggle to build agents. And I see this every day internally at Google, but also externally at Google. And I brought five example on, like, what's really different to how we built traditional software a few years ago and to now how we build agents.

4. 0:44

   And if we, like, on a high level compare them, right? When we wrote software, uh, we created a spec, a PRD, wrote code, sometimes created tests to make sure our code works.

5. 0:56

   We deployed it, and then our user used it. And when building agents, things are a little bit different. Uh, we define instructions on what we want our agent to do.

6. 1:05

   We run it, we observe what it does. We maybe adjust our prompts, maybe we adjust our tools. We run it again, and we have, like, this iterative loop of how can we improve and make our agent way more reliable, which is very different to how we build software.

7. 1:21

   And, like, something I like to compare it to is, like, traditional software is more like we acted as a traffic controller, right? We had control over the streetlights, over how fast you can go, which road you can use, basically how the car drives.

8. 1:34

   And now with agents, we are more of a dispatcher. We tell the agent, "Hey, I want to go to London," and I'm from, like, Germany. I could use the train, I could u-- fly, I could use my car and go, like, under the water.

9. 1:47

   And it's more about, okay, we define the goal on what we want the agent to do, but we don't define the exact step the agent needs to take to achieve that goal.

10. 1:55

    And I mean, every one of you has probably seen in their coding agent that sometimes it does something very weird, but at the end it achieves the outcome, and that's what we want to do.

11. 2:05

    So starting with the first example, text is our new state. I mean, traditionally we had data structures, and everything was kinda mapped to Boolean or to, like, flags we could check.

12. 2:16

    So initially, when we created, for example, a deep research agent, a deep research agent returns a plan to you. "Okay, I'm going to research this and that." In traditional software, we might have had an accept plan, uh, or deny plan, but we couldn't catch semantic meaning.

13. 2:35

    And now what we have with LLMs is they can understand the semantic meaning. So for example, if I have a deep research, um, request on, like, doing some market research, I can approve the initial plan, but I can also on the same time provide additional information.

14. 2:50

    So maybe I want to focus on, like, the US market and ignore California. Uh, maybe I want to provide something additional and not have, like, this multiple steps, right?

15. 2:59

    Traditionally, I would probably said decline, and then it has a follow-up. I m-might need it to provide more input, create a new plan and continue. And another good example is everything related to memory and person-personalization we do cannot really be mapped to data structures, right?

16. 3:16

    The example I here-- I have is, like, um, I'm from Europe, so I mostly use Celsius. But what if I would like to use Fahrenheit for cooking, right? Previously, we might had some flags on, like, the user profile.

17. 3:29

    Is Celsius or is Europe or use Fahrenheit? But I couldn't, like, dynamically adjust based on the user preference, based on what they provide. So really it's all about text and context.

18. 3:42

    I mean, could be images, video or audio as well. But we no longer are really operating in those clear structured data concepts. The other thing is we should start handing over control and the, the trap or the example which we might had from, like, previous customer support is, like, when a user reached out, "Hey, I want to

19. 4:05

    cancel my subscription," I might have had a classification model which kinda classified the intent. Okay, the user wants to churn. And then I had a predefined workflow of, okay, do you try to sell it?

20. 4:18

    Do you cancel the subscription? But there was no, like, dynamic kinda option to, to react to it dynamically. And maybe instead of, like, um, going through the subscription cancel flow, what if your agent, like, kinda tries to understand the meaning and, like, offers something, uh, in-- except to, like, the, the subscription and the user changes their mind

21. 4:41

    and now you have, like, a whole different intent. And it's very hard to model all of those differences and uniqueness and to, to, like, all of those stateful workflows we had before.

22. 4:53

    So we need to, like, trust into the LLM or, like, hand over control that we are no longer working in those purely deterministic, um, environments. The third one is errors are just inputs.

23. 5:08

    So if something in your agent flow fails, we need to treat it as a normal input, as very similar to a user input. In Go, we already do this, right?

24. 5:18

    A function call can be an error or can be a value, and we treat them kinda equally, and we have to do this for agents very similarly. In the past, HTTP requests were very cheap.

25. 5:29

    When some search, some product search failed, you just rerun your request. You redid all of the work, which was okay. But now if you have, like, an agent which takes five minutes, fifteen minutes, and something in the flow breaks and you would start o-all over, you would need to sp- [coughs] You need to spend a lot of compute

26. 5:49

    again to, like, do all of the previous steps, and you also might lose the existing context. So we cannot, like, just start over the whole process. We need to kinda understand and treat errors differently, provide them back to the model, maybe have some other workaround, some additional checks that we-

27. 6:06

    Basically keep going forward in the flow and not like starting over from the beginning.

28. 6:13

    The fourth, uh, example or step is like we need to move from unit tests to evals. So when building software before, right, we wrote integration tests, unit tests, smoke tests, and all kind of different tests.

29. 6:25

    And we assume that when we provide input A for our code B, we will always get C as an output. And that's no longer the case with agent. Agents are non-deterministic.

30. 6:36

    We cannot always guarantee that the same input will lead to the same steps and the same result. So we need to move from unit tests to eval. We need to test how often something works because agents are only successful if they are really reliable, right?

31. 6:51

    If you have a customer agent and the same prompt only works one out of 10 times, it's nothing really you want to put in production, and it becomes very flaky.

32. 7:00

    So we need to test on evals, on how many times it passes. And compared to traditional software, results are very subjective, right? An outcome can be very different if you ask it to create a research report, if you ask it to create a customer feedback kind of scenario.

33. 7:18

    And we need more like qualifying, uh, feedback. Um, LLM-as-a-judge or human expert, for example, is a good way. And we always need to trace what the agent is doing, but we need to grade on the output.

34. 7:29

    We-- Maybe the agent decides for like one user it needs to do like four more steps to do more research than for the other user. It consumes maybe a few more tokens, but at the end, the outcome is really what we need to measure and want to measure in terms of, uh, success.

35. 7:47

    And then the last part is, uh, agents evolve and APIs don't. And if you have worked on the back end, and if you have built an API, you might have seen a lot of methods, API endpoints which feel very self-explaining to you, like delete item feels very self-explaining if you're working on like the product API.

36. 8:10

    But an agent doesn't see the code. An agent doesn't have the context and the background from all those years from you working on the API. So we need to build APIs or tools which are really agent ready, which are self-documenting with semantic interfaces.

37. 8:27

    I would assume if you have like a product microservice and you have a delete item endpoint with an ID, you don't need to like define a docstring what the ID is or what happens if something fails.

38. 8:40

    But our agents only see like the function schemas and the docstrings and the tool definitions. So on the first look, they don't really see what the delete item method does.

39. 8:50

    That's why we need to really adjust to, hey, we need methods, tools which are written for agents to be used and not assume long year developer expertise and people who have built the API.

40. 9:04

    So to, to summarize everything, we need to give trust, but we also verify. We should stop fighting the model. You should not like try to force the model into this one specific workflow with step one, do this, step two, do the other thing.

41. 9:19

    We need to preserve meaning. Um, everything is a context now. We no longer have those very well-defined data structures for all of our applications. We need to design for recovery.

42. 9:31

    Models are not perfect. Agents are not perfect, especially if we have longer running agents. There will be some very weird things happening, so you need to design for recovery.

43. 9:41

    We need to evaluate and don't, don't only assert. Agents are not one hundred percent reliable. We need to find the right balance between how many times our run need to be successful to provide it to the user.

44. 9:53

    And last but not least, build to delete. Um, the bitter lesson is what every one of us is learning is like software is disposable. We are going to rebuild many, many times the same things with better models, better agents.

45. 10:08

    Things will change. And, um, yes, um, it's also available on my blog. So if you want to like look a bit deeper with some code examples. And if not, if you have any questions, feel free to, to reach out to me and perfect on time.

46. 10:22

    Thanks. [clapping] [outro jingle]