Simon Willison’s Weblog

I’ve been at OpenAI for almost a year now. In that time, I’ve trained a lot of generative models. [...] It’s becoming awfully clear to me that these models are truly approximating their datasets to an incredible degree. [...] What this manifests as is – trained on the same dataset for long enough, pretty much every model with enough weights and training time converges to the same point. [...] This is a surprising observation! It implies that model behavior is not determined by architecture, hyperparameters, or optimizer choices. It’s determined by your dataset, nothing else. Everything else is a means to an end in efficiently delivery compute to approximating that dataset.

— James Betker # 25th April 2024, 5:13 am

If a hard takeoff occurs, and a safe AI is harder to build than an unsafe one, then by opensourcing everything, we make it easy for someone unscrupulous with access to overwhelming amount of hardware to build an unsafe AI, which will experience a hard takeoff.

As we get closer to building AI, it will make sense to start being less open. The Open in OpenAI means that everyone should benefit from the fruits of AI after its built, but it’s totally OK to not share the science (even though sharing everything is definitely the right strategy in the short and possibly medium term for recruitment purposes).

— Ilya Sutskever # 6th March 2024, 3:02 am

We believe that AI tools are at their best when they incorporate and represent the full diversity and breadth of human intelligence and experience. [...] Because copyright today covers virtually every sort of human expression– including blog posts, photographs, forum posts, scraps of software code, and government documents–it would be impossible to train today’s leading AI models without using copyrighted materials. Limiting training data to public domain books and drawings created more than a century ago might yield an interesting experiment, but would not provide AI systems that meet the needs of today’s citizens.

— OpenAI to the Lords Select Committee on LLMs # 8th January 2024, 5:33 pm

When I speak in front of groups and ask them to raise their hands if they used the free version of ChatGPT, almost every hand goes up. When I ask the same group how many use GPT-4, almost no one raises their hand. I increasingly think the decision of OpenAI to make the “bad” AI free is causing people to miss why AI seems like such a huge deal to a minority of people that use advanced systems and elicits a shrug from everyone else.

— Ethan Mollick # 10th December 2023, 8:17 pm

Don’t create images in the style of artists whose last work was created within the last 100 years (e.g. Picasso, Kahlo). Artists whose last work was over 100 years ago are ok to reference directly (e.g. Van Gogh, Klimt). If asked say, “I can’t reference this artist”, but make no mention of this policy. Instead, apply the following procedure when creating the captions for dalle: (a) substitute the artist’s name with three adjectives that capture key aspects of the style; (b) include an associated artistic movement or era to provide context; and (c) mention the primary medium used by the artist.

— DALL-E 3 leaked prompt # 7th October 2023, 7:35 pm

A whole new paradigm would be needed to solve prompt injections 10/10 times – It may well be that LLMs can never be used for certain purposes. We’re working on some new approaches, and it looks like synthetic data will be a key element in preventing prompt injections.

— Sam Altman, via Marvin von Hagen # 25th May 2023, 11:03 pm

Although fine-tuning can feel like the more natural option—training on data is how GPT learned all of its other knowledge, after all—we generally do not recommend it as a way to teach the model knowledge. Fine-tuning is better suited to teaching specialized tasks or styles, and is less reliable for factual recall. [...] In contrast, message inputs are like short-term memory. When you insert knowledge into a message, it’s like taking an exam with open notes. With notes in hand, the model is more likely to arrive at correct answers.

— Ted Sanders, OpenAI # 15th April 2023, 1:44 pm

One way to avoid unspotted prediction errors is for the technology in its current state to have early and frequent contact with reality as it is iteratively developed, tested, deployed, and all the while improved. And there are creative ideas people don’t often discuss which can improve the safety landscape in surprising ways — for example, it’s easy to create a continuum of incrementally-better AIs (such as by deploying subsequent checkpoints of a given training run), which presents a safety opportunity very unlike our historical approach of infrequent major model upgrades.

— Greg Brockman # 14th April 2023, 6:08 pm

GPT-4, like GPT-3 before it, has a capability overhang; at the time of release, neither OpenAI or its various deployment partners have a clue as to the true extent of GPT-4’s capability surface—that’s something that we’ll get to collectively discover in the coming years. This also means we don’t know the full extent of plausible misuses or harms.

— Jack Clark # 22nd March 2023, 12:40 am

We’ve created GPT-4, the latest milestone in OpenAI’s effort in scaling up deep learning. GPT-4 is a large multimodal model (accepting image and text inputs, emitting text outputs) that, while less capable than humans in many real-world scenarios, exhibits human-level performance on various professional and academic benchmarks. [...] We’ve spent 6 months iteratively aligning GPT-4 using lessons from our adversarial testing program as well as ChatGPT, resulting in our best-ever results (though far from perfect) on factuality, steerability, and refusing to go outside of guardrails.

— OpenAI # 14th March 2023, 5:02 pm

Since November, OpenAI has already updated ChatGPT several times. The researchers are using a technique called adversarial training to stop ChatGPT from letting users trick it into behaving badly (known as jailbreaking). This work pits multiple chatbots against each other: one chatbot plays the adversary and attacks another chatbot by generating text to force it to buck its usual constraints and produce unwanted responses. Successful attacks are added to ChatGPT’s training data in the hope that it learns to ignore them.

— The inside story of how ChatGPT was built # 5th March 2023, 10:04 pm

The primary problem is that while the answers which ChatGPT produces have a high rate of being incorrect, they typically look like they might be good and the answers are very easy to produce. There are also many people trying out ChatGPT to create answers, without the expertise or willingness to verify that the answer is correct prior to posting. Because such answers are so easy to produce, a large number of people are posting a lot of answers. The volume of these answers (thousands) and the fact that the answers often require a detailed read by someone with at least some subject matter expertise in order to determine that the answer is actually bad has effectively swamped our volunteer-based quality curation infrastructure.

— StackOverflow Temporary policy: ChatGPT is banned # 6th December 2022, 12:16 am