Our alignment research aims to make artificial general intelligence (AGI) aligned with human values and follow human intent. We take an iterative, empirical approach: by attempting to align highly capable AI systems, we can learn what works and what doesn’t, thus refining our ability to make AI systems safer and more aligned. Using scientific experiments, we study how alignment techniques scale and where they will break.

We tackle alignment problems both in our most capable AI systems as well as alignment problems that we expect to encounter on our path to AGI. Our main goal is to push current alignment ideas as far as possible, and to understand and document precisely how they can succeed or why they will fail. We believe that even without fundamentally new alignment ideas, we can likely build sufficiently aligned AI systems to substantially advance alignment research itself.

Unaligned AGI could pose substantial risks to humanity and solving the AGI alignment problem could be so difficult that it will require all of humanity to work together. Therefore we are committed to openly sharing our alignment research when it’s safe to do so: We want to be transparent about how well our alignment techniques actually work in practice and we want every AGI developer to use the world’s best alignment techniques.

At a high-level, our approach to alignment research focuses on engineering a scalable training signal for very smart AI systems that is aligned with human intent. It has three main pillars:

  1. Training AI systems using human feedback
  2. Training AI systems to assist human evaluation
  3. Training AI systems to do alignment research

Aligning AI systems with human values also poses a range of other significant sociotechnical challenges, such as deciding to whom these systems should be aligned. Solving these problems is important to achieving our mission, but we do not discuss them in this post.

Training AI systems using human feedback

RL from human feedback is our main technique for aligning our deployed language models today. We train a class of models called InstructGPT derived from pretrained language models such as GPT-3. These models are trained to follow human intent: both explicit intent given by an instruction as well as implicit intent such as truthfulness, fairness, and safety.

Our results show that there is a lot of low-hanging fruit on alignment-focused fine-tuning right now: InstructGPT is preferred by humans over a 100x larger pretrained model, while its fine-tuning costs <2% of GPT-3’s pretraining compute and about 20,000 hours of human feedback. We hope that our work inspires others in the industry to increase their investment in alignment of large language models and that it raises the bar on users’ expectations about the safety of deployed models.

Our natural language API is a very useful environment for our alignment research: It provides us with a rich feedback loop about how well our alignment techniques actually work in the real world, grounded in a very diverse set of tasks that our customers are willing to pay money for. On average, our customers already prefer to use InstructGPT over our pretrained models.

Yet today’s versions of InstructGPT are quite far from fully aligned: they sometimes fail to follow simple instructions, aren’t always truthful, don’t reliably refuse harmful tasks, and sometimes give biased or toxic responses. Some customers find InstructGPT’s responses significantly less creative than the pretrained models’, something we hadn’t realized from running InstructGPT on publicly available benchmarks. We are also working on developing a more detailed scientific understanding of RL from human feedback and how to improve the quality of human feedback.

Aligning our API is much easier than aligning AGI since most tasks on our API aren’t very hard for humans to supervise and our deployed language models aren’t smarter than humans. We don’t expect RL from human feedback to be sufficient to align AGI, but it is a core building block for the scalable alignment proposals that we’re most excited about, and so it’s valuable to perfect this methodology.

Training models to assist human evaluation

RL from human feedback has a fundamental limitation: it assumes that humans can accurately evaluate the tasks our AI systems are doing. Today humans are pretty good at this, but as models become more capable, they will be able to do tasks that are much harder for humans to evaluate (e.g. finding all the flaws in a large codebase or a scientific paper). Our models might learn to tell our human evaluators what they want to hear instead of telling them the truth. In order to scale alignment, we want to use techniques like recursive reward modeling (RRM), debate, and iterated amplification.

Currently our main direction is based on RRM: we train models that can assist humans at evaluating our models on tasks that are too difficult for humans to evaluate directly. For example:

We trained a model to summarize books. Evaluating book summaries takes a long time for humans if they are unfamiliar with the book, but our model can assist human evaluation by writing chapter summaries. We trained a model to assist humans at evaluating the factual accuracy by browsing the web and providing quotes and links. On simple questions, this model’s outputs are already preferred to responses written by humans. We trained a model to write critical comments on its own outputs: On a query-based summarization task, assistance with critical comments increases the flaws humans find in model outputs by 50% on average. This holds even if we ask humans to write plausible looking but incorrect summaries. We are creating a set of coding tasks selected to be very difficult to evaluate reliably for unassisted humans. We hope to release this data set soon. Our alignment techniques need to work even if our AI systems are proposing very creative solutions (like AlphaGo’s move 37), thus we are especially interested in training models to assist humans to distinguish correct from misleading or deceptive solutions. We believe the best way to learn as much as possible about how to make AI-assisted evaluation work in practice is to build AI assistants.

Training AI systems to do alignment research

There is currently no known indefinitely scalable solution to the alignment problem. As AI progress continues, we expect to encounter a number of new alignment problems that we don’t observe yet in current systems. Some of these problems we anticipate now and some of them will be entirely new.

We believe that finding an indefinitely scalable solution is likely very difficult. Instead, we aim for a more pragmatic approach: building and aligning a system that can make faster and better alignment research progress than humans can.

As we make progress on this, our AI systems can take over more and more of our alignment work and ultimately conceive, implement, study, and develop better alignment techniques than we have now. They will work together with humans to ensure that their own successors are more aligned with humans.

We believe that evaluating alignment research is substantially easier than producing it, especially when provided with evaluation assistance. Therefore human researchers will focus more and more of their effort on reviewing alignment research done by AI systems instead of generating this research by themselves. Our goal is to train models to be so aligned that we can off-load almost all of the cognitive labor required for alignment research.

Importantly, we only need “narrower” AI systems that have human-level capabilities in the relevant domains to do as well as humans on alignment research. We expect these AI systems are easier to align than general-purpose systems or systems much smarter than humans.

Language models are particularly well-suited for automating alignment research because they come “preloaded” with a lot of knowledge and information about human values from reading the internet. Out of the box, they aren’t independent agents and thus don’t pursue their own goals in the world. To do alignment research they don’t need unrestricted access to the internet. Yet a lot of alignment research tasks can be phrased as natural language or coding tasks.

Future versions of WebGPT, InstructGPT, and Codex can provide a foundation as alignment research assistants, but they aren’t sufficiently capable yet. While we don’t know when our models will be capable enough to meaningfully contribute to alignment research, we think it’s important to get started ahead of time. Once we train a model that could be useful, we plan to make it accessible to the external alignment research community.

Limitations

We’re very excited about this approach towards aligning AGI, but we expect that it needs to be adapted and improved as we learn more about how AI technology develops. Our approach also has a number of important limitations:

  • The path laid out here underemphasizes the importance of robustness and interpretability research, two areas OpenAI is currently underinvested in. If this fits your profile, please apply for our research scientist positions!

  • Using AI assistance for evaluation has the potential to scale up or amplify even subtle inconsistencies, biases, or vulnerabilities present in the AI assistant.

  • Aligning AGI likely involves solving very different problems than aligning today’s AI systems. We expect the transition to be somewhat continuous, but if there are major discontinuities or paradigm shifts, then most lessons learned from aligning models like InstructGPT might not be directly useful.

  • The hardest parts of the alignment problem might not be related to engineering a scalable and aligned training signal for our AI systems. Even if this is true, such a training signal will be necessary.

  • It might not be fundamentally easier to align models that can meaningfully accelerate alignment research than it is to align AGI. In other words, the least capable models that can help with alignment research might already be too dangerous if not properly aligned. If this is true, we won’t get much help from our own systems for solving alignment problems.

New Comment
17 comments, sorted by Click to highlight new comments since: Today at 7:53 PM

What's your plan for inner alignment?

Jan Leike has written about inner alignment here https://aligned.substack.com/p/inner-alignment. (I'm at OpenAI, imo I'm not sure if this will work in the worst case and I'm hoping we can come up with a more robust plan)

Yeah, though he hasn't specced out his plan.

I can't speak for OpenAI, but maybe the hope is that we don't need to solve inner alignment in step 1. In step 1 we figure out how to get our narrow-ish, not-yet-superintelligent systems to help us with alignment research even though they aren't fully aligned and can't be trusted to scale up to superintelligence or learn certain dangerous skills. Then in step 2 we solve inner alignment and all remaining alignment problems using the help of those systems.

Interesting idea. I guess that could be worth a shot if we lack anything better.

I don't work for OpenAI. I just saw Sam Altman tweet this post, so I linkposted it here.

Commented in a response to MIRI's A challenge for AGI organizations, and a challenge for readers, along with other people.

Who decides what is harmful, dangerous, or toxic? Do you guys have some official and public guide-lines about that?

So, what's the technical (the one ending with "therefore the probability of disaster is < 1e-9") reason why training InstructGPT was safe?

why training InstructGPT was safe?

Who is claiming that it is safe? I didn't get that implication from the post

"Safe" as in "safe enough for it to be on net better to run it" or "safe enough it wouldn't definitely kill everyone". It's not that I don't have popular intuition that GPT wouldn't kill anyone. It's just that I don't think it's a good habit to run progressively more capable systems while relying on informal intuitions about their safety. And then maybe I will see an explanation for why future safety tools would outpace capability progress, when now we are already at the point where current safety tools are not applicable to current AI systems.

I'm pretty unconvinced by this. I do not think that any substantial fraction of AI x-risk comes from an alignment research who thinks carefully about x-risk deciding that a GPT-3 level system isn't scary enough to take significant precautions with re boxing.

I think taking frivolous risks is bad, but that risk aversion to the point of not being able to pursue otherwise promising research directions seems pretty costly, while the benefits of averting risks >1e-9 is pretty negligible in comparison.

(To be clear, this argument does not apply to more powerful systems! As systems get smarter we should be more capable, and try to be very conservative! But ultimately everything is a trade-off - letting GPT-3 talk to human contractors giving feedback is a way of letting it out of the box!)

I just want the trade-off to be made explicitly. If it turns out that -7 people in expectation is better than thinking about utility functions and all other alternatives - fine. But that's the argument that depends on actual numbers. Yes, it's possible to think informally and correctly. But maybe "an alignment research who thinks carefully about x-risk" wasn't what was happening.

To be clear, this argument does not apply to more powerful systems!

Before running InstructGPT what was the technical reason why it wouldn't be powerful?

While I wouldn't accept that level of risk-aversion, I do agree with a related question: Why do they think they can make significant progress on alignment, exactly?

I mean, I would be glad to hear any number.

Why not? They are running experiments and getting real hands-on experience with AI systems that keep getting better. Seems to me a plausible approach.

There is a school of thought that says you need to mathematically prove that your AGI will be aligned, before you even start building any kind of AI system at all. IMO this would be a great approach if our civilization had strong coordination abilities and unlimited time.