Google Postmortem Tough Love Letter

When I left Google DeepMind last year, I sent this email to much of the organization including Alphabet execs. I am not sure how many read it, but I hoped it would cause at least a little nudge.

It didn’t feel like the right time to share back then, but it’s been awhile and I’m wondering, what do you all think? Have we moved in the right direction since then? What can we do to shift things further?

It appears to be traditional to send a goodbye email when one leaves the company, and upon reflecting on my time here, I felt inspired to write something a little more extensive, with the hope that I will be able to do one last helpful thing while I am here. I wasn’t sure precisely who to send this to, as I am not sure who specifically needs to hear it, and decided to use my best judgement to send it widely, but not to the whole company, as I believe that someone needs to say much of this, and that this is an important opportunity to do so.

My time at Google has been complex, has intertwined in some not so healthy ways with some personal life challenges, and I have now been put in a position where my only prudent choice is to exit the company. Today will be my last day.

First I wish to say that it has been my honor to work with all of you, and I truly believe that you each are brilliant and capable and that you largely have your hearts pointed in the right general direction. I also know that, through all of you, this company has the potential to contribute deep and important technologies that can support the wellbeing of the world. I’ve learned so much from being here, and it has truly been a privilege to be a part of your community.

It feels important that I communicate what I see, which is that this organization strives to be guided by aligned values and principles, and that it really does want to do good for the world, which it often achieves in many beautiful ways. I also see that, like most of us, the company is challenged by its own fears, insecurities, and lack of full commitment to its values, perhaps due to its lack of understanding of how to fully enact them. Some of these lie primarily in the technologies, others in the organization, and yet others in the individual technologists, though really the problem is a complex systems challenge, where the behaviors are locked jointly in all aspects of the system and can only be changed by addressing each opportunity to improve.

I believe that the technologies you are developing will change the face of our Earth system and that it is up to you to create them in a way where they support a healthy, happy, and harmonious world rather than one that is sick, sad, and fragmented. I believe that this can only be done in an environment where we recognize that all of us Earth dwellers are in this together, and where we care for each other and support each other to the same standard we have for ourselves. I hope I can leave you with some of my high level thoughts about how to do this at Google.

I have much to say, but I want to keep it (relatively) short and sweet. Here are five wishes that I have for the organization:

1. Let us not just “do no evil”, let us do good. (I recognize that this is an outdated informal motto, but it does often seem to be how decisions are made around here...) It is not enough to mitigate harms of the kinds that we recognize, categorize, and measure. Many other kinds of harm exist, and we cannot catch them all with this approach. Instead of moving away from harm, I hope that you all can learn to focus on moving towards supporting wellbeing for everyone and everything affected by your work, whether it be users, employees, communities, other companies, or our natural environment. Responsible AI is not just about mitigating harm, but about supporting wellbeing. Trust in the importance of enacting good, and you will receive everything you need and more, though sometimes in forms that require adaptation to recognize.

2. Incentivize good for the world. If the stewards of world-changing technologies are not listened to, respected, or given space to do what they believe to be best for the world, then they will have to choose between doing what is best for the world and what is best for themselves (e.g. keeping their job, getting a promotion, having enough time at home to take good care of themselves and their families, etc.). When given this choice, depending on the parameters, even well-meaning people will sacrifice bits of the world for themselves, if only because, if they do not take care of themselves, no one else will. In certain real circumstances, this can cause significant material harm to the world. Instead, make sure that employees are supported in getting everything that they need, and align incentive structures towards what you believe to be best for the long-term holistic future of our Earth (including of course what keeps Google sustained and becoming a more effective agent for good).

3. Listen to other employees, trust them, and treat them as unique individuals. These values are all articulated in our policies, but I often see them directly contradicted in the policies and practices of the organization and by the actions of individuals. When others around you tell you what they believe to be best, could you take the time to listen openly, and when there is an opportunity, trust that maybe they see something that you don’t and give them space to show you what they mean? When they tell you honestly and directly what they need to be well, is it right to ask only what you are supposed to do according to company policy and look for the easiest answer? Wouldn’t it be better to try your hardest to figure out how you can support them given the policies, and how you might be able to influence change in the organization such that the needed support is more easeful for others in the future? If someone is (or a collection of people are) robustly unwell, then something is broken, and if something is broken, it really can be fixed. Is it not inconsistent with your values to bucket and categorize and otherwise reductively simplify people? Wouldn’t it be healthier to work towards processes and systems that are abstracted in the right ways to allow you to treat them as individuals?

4. Invest in foundational understanding. Don’t be afraid to move slower sometimes in order to move fast. In order to build more powerful and general technologies, we often need to build from more general and foundational understanding. Reinvest in this understanding, in basic research, and in the freedom and space required to make this possible. Foundational research is the reason why we are here now, and let us not forget, even when things are moving quickly due to a paradigm shift, that progress is a stochastic multiple of effort and understanding, and that investing in deeper understanding, supports deeper progress.

5. Do not measure your performance primarily against others; lead the way. I often hear in meetings that we are comparing our performance against our peers and that we feel we need to race them and do better than them. To a certain extent this is an inevitable outcome of market forces that emerge from our present economic paradigm, but, while admittedly useful, it is ultimately a myopic way to measure progress. Take time and space to envision the world that you wish to see, a world that is better for all beings in some way, and mark progress against this as a primary metric instead. This should apply holistically, as much to business metrics as to model and product evaluations, as to how you treat your employees, and how you communicate with the world. For example, justifying things in terms of what is “industry standard” has to stop. You are Google, and while you do have real constraints, you also have immense resources, talent, and freedom with which to shift the company and the world in a direction that is most beautiful and supportive of the wellbeing of all of the parts that compose our Earth. Don’t do what is industry standard. Do what is holistically best.

I do recognize that the company is, in some ways, already moving in these directions, that this is challenging, and that it takes time. But people are still getting hurt and I hope that having some clarity will help orient you towards improvement sooner rather than later. I am also not measuring any of this against other peer organizations, whom I believe you are likely exceeding in many of these ways. I suggest measuring instead by what is needed for you all to become the most responsible versions of the world leaders that you are increasingly becoming.

Finally, while you all are doing truly magnificent and incredible work in many ways, I do not hear a lot of laughs, smiles, and deep relaxed breaths in the offices these days, and I believe this to be the most important measure of the wellbeing of a community. Strive to help people to feel lighter, more happy, and more free. If you lead with your heart, and care for each other, then it will feel better for each of you, and all this joy and perhaps even love will flow from your lives into the lives of all of the beings you touch with your work, and you can help lead the way to a world that is well.

I wish you all the best and am so grateful to have been a part of your organization. If you would like to discuss any of these ideas further, feel free to write to me at ___@gmail.com.

love,

larry muhlstein