Oxford Mathematics Public LectureDeep Maths - Machine Learning and MathematicsIn December 2021 mathematicians at Oxford and Sydney universities together with their collaborators at DeepMind announced that they had successfully used tools from machine learning to discover new patterns in mathematics. But what exactly had they done and what are its implications for the future of mathematics and mathematicians?This online event will feature short talks from each of the four collaborators, explaining their work, followed by a panel discussion addressing its wider implications. If you wish to submit a question, please email external-relations@maths.ox.ac.ukThe speakers:Alex Davies - DeepMindAndras Juhasz - University of OxfordMarc Lackenby - University of OxfordGeordie Williamson - University of SydneyThe panel will be chaired by Jon Keating, Sedleian Professor of Natural Philosophy in Oxford.This is an online only lecture which every one is free to watch: www.youtube.com/c/OxfordMathematicsThe Oxford Mathematics Public Lectures are generously supported by XTX Markets.

New, state-of-the-art language models like DeepMind's Gopher, Microsoft's Megatron, and OpenAI's GPT-3 are driving a wave of innovation in NLP. How do you measure the performance of these language models to see how good they are? In a previous post, we gave an overview of different language model evaluation metrics. This post dives more deeply into one of the most popular: a metric known as perplexity. Surge AI delivers better data, faster.

GPT-1 was released on June 11, 2018. When this model was released by OpenAI, there was much excitement. It was the transformer structure combined with unsupervised pre-training with promising results. The key difference between GPT-1 and the other language-based models before it, is that it was fine-tuned, or trained for specific tasks. GPT-2 was introduced in February 2019.

The deployment of powerful AI systems has enriched our understanding of safety and misuse far more than would have been possible through research alone. Here, we describe our latest thinking in the hope of helping other AI developers address safety and misuse of deployed models. Over the past two years, we've learned a lot about how language models can be used and abused--insights we couldn't have gained without the experience of real-world deployment. In June 2020, we began giving access to developers and researchers to the OpenAI API, an interface for accessing and building applications on top of new AI models developed by OpenAI. Deploying GPT-3, Codex, and other models in a way that reduces risks of harm has posed various technical and policy challenges.

The Internet of the future might be something completely different from what we know currently. But this change will be for the better good or worse. Experts from Copenhagen Institute for Future Studies (CIFS) have raised questions about AI-generated content, and how it might rule digital locations and the much-hyped metaverse. According to CIFS expert Timothy Shoup, 99% or more of the internet's content will be generated by artificial intelligence by 2025 to 2030, especially if models such as OpenAI's GPT-3 witness a wider use.

The detection of computer-generated text is an area of rapidly increasing significance as nascent generative models allow for efficient creation of compelling human-like text, which may be abused for the purposes of spam, disinformation, phishing, or online influence campaigns. Past work has studied detection of current state-of-the-art models, but despite a developing threat landscape, there has been minimal analysis of the robustness of detection methods to adversarial attacks. To this end, we evaluate neural and non-neural approaches on their ability to detect computer-generated text, their robustness against text adversarial attacks, and the impact that successful adversarial attacks have on human judgement of text quality. We find that while statistical features underperform neural features, statistical features provide additional adversarial robustness that can be leveraged in ensemble detection models. In the process, we find that previously effective complex phrasal features for detection of computer-generated text hold little predictive power against contemporary generative models, and identify promising statistical features to use instead. Finally, we pioneer the usage of $\Delta$MAUVE as a proxy measure for human judgement of adversarial text quality.

According to CIFS expert Timothy Shoup, 99% or more of the internet's content will be generated by artificial intelligence by 2025 to 2030, especially if models such as OpenAI's GPT-3 witness a wider use. "The internet would be completely unrecognizable," Shoup told colleague Sofie Hvitved. As the capabilities of AI advance, it could start creating entire online worlds, alongside all the things that inhabit them. This will also include all the online material that humans make use of even now. This could give birth to things that we could only imagine right now.

Abductive reasoning, frequently misidentified as deductive reasoning, is the process of making a plausible prediction when faced with incomplete information. For example, given a photo showing a toppled truck and a police cruiser on a snowy freeway, abductive reasoning may lead someone to infer that dangerous road conditions caused an accident. Humans can quickly consider this sort of context to arrive at a hypothesis. But AI struggles, despite recent technical advances. Motivated to explore the challenge, researchers at the Allen Institute for Artificial Intelligence, the University of California, Berkeley, and the MIT-IBM Watson AI lab created a dataset called Sherlock, a collection of over 100,000 images of scenes paired with clues a viewer could use to answer questions about the scenes.

Controlling nuclear fusion on Earth is hard, however. The problem is that atomic nuclei repel each other. Smashing them together inside a reactor can only be done at extremely high temperatures, often reaching hundreds of millions of degrees--hotter than the center of the sun. At these temperatures, matter is neither solid, liquid, nor gas. It enters a fourth state, known as plasma: a roiling, superheated soup of particles.

Scientists at EPFL's Swiss Plasma Center and DeepMind have jointly developed a new method for controlling plasma configurations for use in nuclear fusion research. EPFL's Swiss Plasma Center (SPC) has decades of experience in plasma physics and plasma control methods. DeepMind is a scientific discovery company acquired by Google in 2014 that's committed to'solving intelligence to advance science and humanity. Together, they have developed a new magnetic control method for plasmas based on deep reinforcement learning, and applied it to a real-world plasma for the first time in the SPC's tokamak research facility, TCV. Their study has just been published in Nature.