# 2017 in review

|   |  MIRI Strategy

This post reviews MIRI’s activities in 2017, including research, recruiting, exposition, and fundraising activities.

2017 was a big transitional year for MIRI, as we took on new research projects that have a much greater reliance on hands-on programming work and experimentation. We’ve continued these projects in 2018, and they’re described more in our 2018 update. This meant a major focus on laying groundwork for much faster growth than we’ve had in the past, including setting up infrastructure and changing how we recruit to reach out to more people with engineering backgrounds.

# MIRI’s newest recruit: Edward Kmett!

|   |  News

Prolific Haskell developer Edward Kmett has joined the MIRI team!

Edward is perhaps best known for popularizing the use of lenses for functional programming. Lenses are a tool that provides a compositional vocabulary for accessing parts of larger structures and describing what you want to do with those parts.

Beyond the lens library, Edward maintains a significant chunk of all libraries around the Haskell core libraries, covering everything from automatic differentiation (used heavily in deep learning, computer vision, and financial risk) to category theory (biased heavily towards organizing software) to graphics, SAT bindings, RCU schemes, tools for writing compilers, and more.

Initial support for Edward joining MIRI is coming in the form of funding from long-time MIRI donor Jaan Tallinn. Increased donor enthusiasm has put MIRI in a great position to take on more engineers in general, and to consider highly competitive salaries for top-of-their-field engineers like Edward who are interested in working with us.

At MIRI, Edward is splitting his time between helping us grow our research team and diving in on a line of research he’s been independently developing in the background for some time: building a new language and infrastructure to make it easier for people to write highly complex computer programs with known desirable properties. While we are big fans of his work, Edward’s research is independent of the directions we described in our 2018 Update, and we don’t consider it part of our core research focus.

We’re hugely excited to have Edward at MIRI. We expect to learn and gain a lot from our interactions, and we also hope that having Edward on the team will let him and other MIRI staff steal each other’s best problem-solving heuristics and converge on research directions over time.

As described in our recent update, our new lines of research are heavy on the mix of theoretical rigor and hands-on engineering that Edward and the functional programming community are well-known for:

In common between all our new approaches is a focus on using high-level theoretical abstractions to enable coherent reasoning about the systems we build. A concrete implication of this is that we write lots of our code in Haskell, and are often thinking about our code through the lens of type theory.

MIRI’s nonprofit mission is to ensure that smarter-than-human AI systems, once developed, have a positive impact on the world. And we want to actually succeed in that goal, not just go through the motions of working on the problem.

Our current model of the challenges involved says that the central sticking point for future engineers will likely be that the building blocks of AI just aren’t sufficiently transparent. We think that someone, somewhere, needs to develop some new foundations and deep theory/insights, above and beyond what’s likely to arise from refining or scaling up currently standard techniques.

We think that the skillset of functional programmers tends to be particularly well-suited to this kind of work, and we believe that our new research areas can absorb a large number of programmers and computer scientists. So we want this hiring announcement to double as a hiring pitch: consider joining our research effort!

To learn more about what it’s like to work at MIRI and what kinds of candidates we’re looking for, see our last big post, or shoot MIRI researcher Buck Shlegeris an email.

# MIRI’s 2018 Fundraiser

|   |  News

Update January 2019: MIRI’s 2018 fundraiser is now concluded.

$946,981 | |$0

|
$300,000 |$600,000

|
$900,000 |$1,200,000

### Fundraiser concluded

##### 345 donors contributed

MIRI is a math/CS research nonprofit with a mission of maximizing the potential humanitarian benefit of smarter-than-human artificial intelligence. You can learn more about the kind of work we do in “Ensuring Smarter-Than-Human Intelligence Has A Positive Outcome” and “Embedded Agency.”

Our funding targets this year are based on a goal of raising enough in 2018 to match our “business-as-usual” budget next year. We view “make enough each year to pay for the next year” as a good heuristic for MIRI, given that we’re a quickly growing nonprofit with a healthy level of reserves and a budget dominated by researcher salaries.

# 2018 Update: Our New Research Directions

|   |  MIRI Strategy, News

For many years, MIRI’s goal has been to resolve enough fundamental confusions around alignment and intelligence to enable humanity to think clearly about technical AI safety risks—and to do this before this technology advances to the point of potential catastrophe. This goal has always seemed to us to be difficult, but possible.1

Last year, we said that we were beginning a new research program aimed at this goal.2 Here, we’re going to provide background on how we’re thinking about this new set of research directions, lay out some of the thinking behind our recent decision to do less default sharing of our research, and make the case for interested software engineers to join our team and help push our understanding forward.

1. This post is an amalgam put together by a variety of MIRI staff. The byline saying “Nate” means that I (Nate) endorse the post, and that many of the concepts and themes come in large part from me, and I wrote a decent number of the words. However, I did not write all of the words, and the concepts and themes were built in collaboration with a bunch of other MIRI staff. (This is roughly what bylines have meant on the MIRI blog for a while now, and it’s worth noting explicitly.)
2. See our 2017 strategic update and fundraiser posts for more details.

# Embedded Curiosities

|   |  Analysis

This is the conclusion of the Embedded Agency series. Previous posts:

A final word on curiosity, and intellectual puzzles:

I described an embedded agent, Emmy, and said that I don’t understand how she evaluates her options, models the world, models herself, or decomposes and solves problems.

In the past, when researchers have talked about motivations for working on problems like these, they’ve generally focused on the motivation from AI risk. AI researchers want to build machines that can solve problems in the general-purpose fashion of a human, and dualism is not a realistic framework for thinking about such systems. In particular, it’s an approximation that’s especially prone to breaking down as AI systems get smarter. When people figure out how to build general AI systems, we want those researchers to be in a better position to understand their systems, analyze their internal properties, and be confident in their future behavior.

This is the motivation for most researchers today who are working on things like updateless decision theory and subsystem alignment. We care about basic conceptual puzzles which we think we need to figure out in order to achieve confidence in future AI systems, and not have to rely quite so much on brute-force search or trial and error.

But the arguments for why we may or may not need particular conceptual insights in AI are pretty long. I haven’t tried to wade into the details of that debate here. Instead, I’ve been discussing a particular set of research directions as an intellectual puzzle, and not as an instrumental strategy.

One downside of discussing these problems as instrumental strategies is that it can lead to some misunderstandings about why we think this kind of work is so important. With the “instrumental strategies” lens, it’s tempting to draw a direct line from a given research problem to a given safety concern. But it’s not that I’m imagining real-world embedded systems being “too Bayesian” and this somehow causing problems, if we don’t figure out what’s wrong with current models of rational agency. It’s certainly not that I’m imagining future AI systems being written in second-order logic! In most cases, I’m not trying at all to draw direct lines between research problems and specific AI failure modes.

What I’m instead thinking about is this: We sure do seem to be working with the wrong basic concepts today when we try to think about what agency is, as seen by the fact that these concepts don’t transfer well to the more realistic embedded framework.

If AI developers in the future are still working with these confused and incomplete basic concepts as they try to actually build powerful real-world optimizers, that seems like a bad position to be in. And it seems like the research community is unlikely to figure most of this out by default in the course of just trying to develop more capable systems. Evolution certainly figured out how to build human brains without “understanding” any of this, via brute-force search.

Embedded agency is my way of trying to point at what I think is a very important and central place where I feel confused, and where I think future researchers risk running into confusions too.

There’s also a lot of excellent AI alignment research that’s being done with an eye toward more direct applications; but I think of that safety research as having a different type signature than the puzzles I’ve talked about here.

Intellectual curiosity isn’t the ultimate reason we privilege these research directions. But there are some practical advantages to orienting toward research questions from a place of curiosity at times, as opposed to only applying the “practical impact” lens to how we think about the world.

When we apply the curiosity lens to the world, we orient toward the sources of confusion preventing us from seeing clearly; the blank spots in our map, the flaws in our lens. It encourages re-checking assumptions and attending to blind spots, which is helpful as a psychological counterpoint to our “instrumental strategy” lens—the latter being more vulnerable to the urge to lean on whatever shaky premises we have on hand so we can get to more solidity and closure in our early thinking.

Embedded agency is an organizing theme behind most, if not all, of our big curiosities. It seems like a central mystery underlying many concrete difficulties.

# Subsystem Alignment

|   |  Analysis

You want to figure something out, but you don’t know how to do that yet.

You have to somehow break up the task into sub-computations. There is no atomic act of “thinking”; intelligence must be built up of non-intelligent parts.

The agent being made of parts is part of what made counterfactuals hard, since the agent may have to reason about impossible configurations of those parts.

Being made of parts is what makes self-reasoning and self-modification even possible.

What we’re primarily going to discuss in this section, though, is another problem: when the agent is made of parts, there could be adversaries not just in the external environment, but inside the agent as well.

This cluster of problems is Subsystem Alignment: ensuring that subsystems are not working at cross purposes; avoiding subprocesses optimizing for unintended goals.

• benign induction
• benign optimization
• transparency
• mesa-optimizers

# Robust Delegation

|   |  Analysis

Because the world is big, the agent as it is may be inadequate to accomplish its goals, including in its ability to think.

Because the agent is made of parts, it can improve itself and become more capable.

Improvements can take many forms: The agent can make tools, the agent can make successor agents, or the agent can just learn and grow over time. However, the successors or tools need to be more capable for this to be worthwhile.

This gives rise to a special type of principal/agent problem:

You have an initial agent, and a successor agent. The initial agent gets to decide exactly what the successor agent looks like. The successor agent, however, is much more intelligent and powerful than the initial agent. We want to know how to have the successor agent robustly optimize the initial agent’s goals.

Here are three examples of forms this principal/agent problem can take:

In the AI alignment problem, a human is trying to build an AI system which can be trusted to help with the human’s goals.

In the tiling agents problem, an agent is trying to make sure it can trust its future selves to help with its own goals.

Or we can consider a harder version of the tiling problem—stable self-improvement—where an AI system has to build a successor which is more intelligent than itself, while still being trustworthy and helpful.

For a human analogy which involves no AI, you can think about the problem of succession in royalty, or more generally the problem of setting up organizations to achieve desired goals without losing sight of their purpose over time.

The difficulty seems to be twofold:

First, a human or AI agent may not fully understand itself and its own goals. If an agent can’t write out what it wants in exact detail, that makes it hard for it to guarantee that its successor will robustly help with the goal.

Second, the idea behind delegating work is that you not have to do all the work yourself. You want the successor to be able to act with some degree of autonomy, including learning new things that you don’t know, and wielding new skills and capabilities.

In the limit, a really good formal account of robust delegation should be able to handle arbitrarily capable successors without throwing up any errors—like a human or AI building an unbelievably smart AI, or like an agent that just keeps learning and growing for so many years that it ends up much smarter than its past self.

The problem is not (just) that the successor agent might be malicious. The problem is that we don’t even know what it means not to be.

This problem seems hard from both points of view.

The initial agent needs to figure out how reliable and trustworthy something more powerful than it is, which seems very hard. But the successor agent has to figure out what to do in situations that the initial agent can’t even understand, and try to respect the goals of something that the successor can see is inconsistent, which also seems very hard.

At first, this may look like a less fundamental problem than “make decisions” or “have models”. But the view on which there are multiple forms of the “build a successor” problem is itself a dualistic view.

To an embedded agent, the future self is not privileged; it is just another part of the environment. There isn’t a deep difference between building a successor that shares your goals, and just making sure your own goals stay the same over time.

So, although I talk about “initial” and “successor” agents, remember that this isn’t just about the narrow problem humans currently face of aiming a successor. This is about the fundamental problem of being an agent that persists and learns over time.

We call this cluster of problems Robust Delegation. Examples include: