The public release of ChatGPT has resulted in considerable publicity and has led to wide-spread discussion of the usefulness and capabilities of generative AI language models. Its ability to extract and summarise data from textual sources and present them as human-like contextual responses makes it an eminently suitable tool to answer questions users might ask. This paper tested what archaeological literature appears to have been included in ChatGPT's training phase. While ChatGPT offered seemingly pertinent references, a large percentage proved to be fictitious. Using cloze analysis to make inferences on the sources 'memorised' by a generative AI model, this paper was unable to prove that ChatGPT had access to the full texts of the genuine references. It can be shown that all references provided by ChatGPT that were found to be genuine have also been cited on Wikipedia pages. This strongly indicates that the source base for at least some of the data is found in those pages. The implications of this in relation to data quality are discussed.

In real-world scenarios, human dialogues are multi-round and diverse. Furthermore, human instructions can be unclear and human responses are unrestricted. Interactive robots face difficulties in understanding human intents and generating suitable strategies for assisting individuals through manipulation. In this article, we propose Mani-GPT, a Generative Pre-trained Transformer (GPT) for interactive robotic manipulation. The proposed model has the ability to understand the environment through object information, understand human intent through dialogues, generate natural language responses to human input, and generate appropriate manipulation plans to assist the human. This makes the human-robot interaction more natural and humanized. In our experiment, Mani-GPT outperforms existing algorithms with an accuracy of 84.6% in intent recognition and decision-making for actions. Furthermore, it demonstrates satisfying performance in real-world dialogue tests with users, achieving an average response accuracy of 70%.

We introduce OpenFlamingo, a family of autoregressive vision-language models ranging from 3B to 9B parameters. OpenFlamingo is an ongoing effort to produce an open-source replication of DeepMind's Flamingo models. On seven vision-language datasets, OpenFlamingo models average between 80 - 89% of corresponding Flamingo performance. This technical report describes our models, training data, hyperparameters, and evaluation suite. We share our models and code at https://github.com/mlfoundations/open_flamingo.

Large language models (LLMs) offer the potential to automate data science through natural language interfaces, but it is difficult to embed complex models or datasets in confined context windows. While GPT-4 has a context window size of up to 32k tokens, paying equal attention to all parts of the context remains a challenge [1] and the practicality of lengthy context windows is questionable. Machine learning models often involve billions of parameters, accentuating the need for compact, modular function representations that more easily interface with LLMs. In this paper, we show that LLMs pair remarkably well with interpretable models that are decomposable into modular components. Specifically, we show that GPT-4 is able to describe, interpret and debug univariate graphs, and by applying a form of chain-of-thought reasoning[2], GPT-4 can understand Generalized Additive Models (GAMs). GAMs [3, 4] represent complex outcomes as sums of univariate component functions (graphs); thus, by analyzing each of these component functions in turn, the LLM does not need to understand the entire model at once. After analyzing and summarizing each graph, the LLM can operate on component summaries to produce model-level analyses. This modularity simplifies the application of LLMs to data science and machine learning and enables LLM-based analyses to scale to very large datasets while staying within small context windows.

Tables are prevalent in real-world databases, requiring significant time and effort for humans to analyze and manipulate. The advancements in large language models (LLMs) have made it possible to interact with tables using natural language input, bringing this capability closer to reality. In this paper, we present TableGPT, a unified fine-tuned framework that enables LLMs to understand and operate on tables using external functional commands. It introduces the capability to seamlessly interact with tables, enabling a wide range of functionalities such as question answering, data manipulation (e.g., insert, delete, query, and modify operations), data visualization, analysis report generation, and automated prediction. TableGPT aims to provide convenience and accessibility to users by empowering them to effortlessly leverage tabular data. At the core of TableGPT lies the novel concept of global tabular representations, which empowers LLMs to gain a comprehensive understanding of the entire table beyond meta-information. By jointly training LLMs on both table and text modalities, TableGPT achieves a deep understanding of tabular data and the ability to perform complex operations on tables through chain-of-command instructions. Importantly, TableGPT offers the advantage of being a self-contained system rather than relying on external API interfaces. Moreover, it supports efficient data process flow, query rejection (when appropriate) and private deployment, enabling faster domain data fine-tuning and ensuring data privacy, which enhances the framework's adaptability to specific use cases.

Adaptive gradient methods, such as Adam and LAMB, have demonstrated excellent performance in the training of large language models. Nevertheless, the need for adaptivity requires maintaining second-moment estimates of the per-parameter gradients, which entails a high cost of extra memory overheads. To solve this problem, several memory-efficient optimizers (e.g., Adafactor) have been proposed to obtain a drastic reduction in auxiliary memory usage, but with a performance penalty. In this paper, we first study a confidence-guided strategy to reduce the instability of existing memory efficient optimizers. Based on this strategy, we propose CAME to simultaneously achieve two goals: fast convergence as in traditional adaptive methods, and low memory usage as in memory-efficient methods. Extensive experiments demonstrate the training stability and superior performance of CAME across various NLP tasks such as BERT and GPT-2 training. Notably, for BERT pre-training on the large batch size of 32,768, our proposed optimizer attains faster convergence and higher accuracy compared with the Adam optimizer. The implementation of CAME is publicly available.

The rise of powerful large language models (LLMs) brings about tremendous opportunities for innovation but also looming risks for individuals and society at large. We have reached a pivotal moment for ensuring that LLMs and LLM-infused applications are developed and deployed responsibly. However, a central pillar of responsible AI -- transparency -- is largely missing from the current discourse around LLMs. It is paramount to pursue new approaches to provide transparency for LLMs, and years of research at the intersection of AI and human-computer interaction (HCI) highlight that we must do so with a human-centered perspective: Transparency is fundamentally about supporting appropriate human understanding, and this understanding is sought by different stakeholders with different goals in different contexts. In this new era of LLMs, we must develop and design approaches to transparency by considering the needs of stakeholders in the emerging LLM ecosystem, the novel types of LLM-infused applications being built, and the new usage patterns and challenges around LLMs, all while building on lessons learned about how people process, interact with, and make use of information. We reflect on the unique challenges that arise in providing transparency for LLMs, along with lessons learned from HCI and responsible AI research that has taken a human-centered perspective on AI transparency. We then lay out four common approaches that the community has taken to achieve transparency -- model reporting, publishing evaluation results, providing explanations, and communicating uncertainty -- and call out open questions around how these approaches may or may not be applied to LLMs. We hope this provides a starting point for discussion and a useful roadmap for future research.

Membership Inference attacks (MIAs) aim to predict whether a data sample was present in the training data of a machine learning model or not, and are widely used for assessing the privacy risks of language models. Most existing attacks rely on the observation that models tend to assign higher probabilities to their training samples than non-training points. However, simple thresholding of the model score in isolation tends to lead to high false-positive rates as it does not account for the intrinsic complexity of a sample. Recent work has demonstrated that reference-based attacks which compare model scores to those obtained from a reference model trained on similar data can substantially improve the performance of MIAs. However, in order to train reference models, attacks of this kind make the strong and arguably unrealistic assumption that an adversary has access to samples closely resembling the original training data. Therefore, we investigate their performance in more realistic scenarios and find that they are highly fragile in relation to the data distribution used to train reference models. To investigate whether this fragility provides a layer of safety, we propose and evaluate neighbourhood attacks, which compare model scores for a given sample to scores of synthetically generated neighbour texts and therefore eliminate the need for access to the training data distribution. We show that, in addition to being competitive with reference-based attacks that have perfect knowledge about the training data distribution, our attack clearly outperforms existing reference-free attacks as well as reference-based attacks with imperfect knowledge, which demonstrates the need for a reevaluation of the threat model of adversarial attacks.

Numerous parties are calling for the democratisation of AI, but the phrase is used to refer to a variety of goals, the pursuit of which sometimes conflict. This paper identifies four kinds of AI democratisation that are commonly discussed: (1) the democratisation of AI use, (2) the democratisation of AI development, (3) the democratisation of AI profits, and (4) the democratisation of AI governance. Numerous goals and methods of achieving each form of democratisation are discussed. The main takeaway from this paper is that AI democratisation is a multifarious and sometimes conflicting concept that should not be conflated with improving AI accessibility. If we want to move beyond ambiguous commitments to democratising AI, to productive discussions of concrete policies and trade-offs, then we need to recognise the principal role of the democratisation of AI governance in navigating tradeoffs and risks across decisions around use, development, and profits.

We present a critical assessment of Piantadosi's (2023) claim that "Modern language models refute Chomsky's approach to language," focusing on four main points. First, despite the impressive performance and utility of large language models (LLMs), humans achieve their capacity for language after exposure to several orders of magnitude less data. The fact that young children become competent, fluent speakers of their native languages with relatively little exposure to them is the central mystery of language learning to which Chomsky initially drew attention, and LLMs currently show little promise of solving this mystery. Second, what can the artificial reveal about the natural? Put simply, the implications of LLMs for our understanding of the cognitive structures and mechanisms underlying language and its acquisition are like the implications of airplanes for understanding how birds fly. Third, LLMs cannot constitute scientific theories of language for several reasons, not least of which is that scientific theories must provide interpretable explanations, not just predictions. This leads to our final point: to even determine whether the linguistic and cognitive capabilities of LLMs rival those of humans requires explicating what humans' capacities actually are. In other words, it requires a separate theory of language and cognition; generative linguistics provides precisely such a theory. As such, we conclude that generative linguistics as a scientific discipline will remain indispensable throughout the 21st century and beyond.