Last March, just two weeks after GPT-4 was released, researchers at Microsoft quietly announced a plan to compile millions of APIs—tools that can do everything from ordering a pizza to solving physics equations to controlling the TV in your living room—into a compendium that would be made accessible to large language models (LLMs). This was just one milestone in the race across industry and academia to find the best ways to teach LLMs how to manipulate tools, which would supercharge the potential of AI more than any of the impressive advancements we’ve seen to date.
The Microsoft project aims to teach AI how to use any and all digital tools in one fell swoop, a clever and efficient approach. Today, LLMs can do a pretty good job of recommending pizza toppings to you if you describe your dietary preferences and can draft dialog that you could use when you call the restaurant. But most AI tools can’t place the order, not even online. In contrast, Google’s seven-year-old Assistant tool can synthesize a voice on the telephone and fill out an online order form, but it can’t pick a restaurant or guess your order. By combining these capabilities, though, a tool-using AI could do it all. An LLM with access to your past conversations and tools like calorie calculators, a restaurant menu database, and your digital payment wallet could feasibly judge that you are trying to lose weight and want a low-calorie option, find the nearest restaurant with toppings you like, and place the delivery order. If it has access to your payment history, it could even guess at how generously you usually tip. If it has access to the sensors on your smartwatch or fitness tracker, it might be able to sense when your blood sugar is low and order the pie before you even realize you’re hungry.
Perhaps the most compelling potential applications of tool use are those that give AIs the ability to improve themselves. — Read More
Tag Archives: Multi-Modal
ImageBind: One Embedding Space To Bind Them All
We present ImageBind, an approach to learn a joint embedding across six different modalities – images, text, audio, depth, thermal, and IMU data. We show that all combinations of paired data are not necessary to train such a joint embedding, and only image-paired data is sufficient to bind the modalities together. ImageBind can leverage recent large scale vision-language models, and extends their zero-shot capabilities to new modalities just by using their natural pairing with images. It enables novel emergent applications ‘out-of-the-box’ including cross-modal retrieval, composing modalities with arithmetic, cross-modal detection and generation. The emergent capabilities improve with the strength of the image encoder and we set a new state-of-the-art on emergent zero-shot recognition tasks across modalities, outperforming specialist supervised models. Finally, we show strong few-shot recognition results outperforming prior work, and that ImageBind serves as a new way to evaluate vision models for visual and non-visual tasks. — Read More
gpt-author
This project utilizes a chain of GPT-4 and Stable Diffusion API calls to generate an original fantasy novel. Users can provide an initial prompt and enter how many chapters they’d like it to be, and the AI then generates an entire novel, outputting an EPUB file compatible with e-book readers.
A 15-chapter novel can cost as little as $4 to produce, and is written in just a few minutes. — Read More
Generating Images with Multimodal Language Models
We propose a method to fuse frozen text-only large language models (LLMs) with pre-trained image encoder and decoder models, by mapping between their embedding spaces. Our model demonstrates a wide suite of multimodal capabilities: image retrieval, novel image generation, and multimodal dialogue. Ours is the first approach capable of conditioning on arbitrarily interleaved image and text inputs to generate coherent image (and text) outputs. To achieve strong performance on image generation, we propose an efficient mapping network to ground the LLM to an off-the-shelf text-to-image generation model. This mapping network translates hidden representations of text into the embedding space of the visual models, enabling us to leverage the strong text representations of the LLM for visual outputs. Our approach outperforms baseline generation models on tasks with longer and more complex language. In addition to novel image generation, our model is also capable of image retrieval from a prespecified dataset, and decides whether to retrieve or generate at inference time. This is done with a learnt decision module which conditions on the hidden representations of the LLM. Our model exhibits a wider range of capabilities compared to prior multimodal language models. It can process image-and-text inputs, and produce retrieved images, generated images, and generated text — outperforming non-LLM based generation models across several text-to-image tasks that measure context dependence. — Read More
Brain Map
Functional magnetic resonance imaging (fMRI) was used to measure brain activity in seven people while they listened to more than 2 hours of stories from The Moth Radio Hour. This data was used to estimate voxel-wise models that predict brain activity in each voxel (volumetric pixel) based on the meaning of the words in the stories. Read the paper describing this research here.
This site provides an interactive 3D viewer for models fit to one subject’s brain. Read More
WenLan: Bridging Vision and Language by Large-Scale Multi-Modal Pre-Training
Multi-modal pre-training models have been intensively explored to bridge vision and language in recent years. However, most of them explicitly model the cross-modal interaction between image-text pairs, by assuming that there exists strong semantic correlation between the text and image modalities. Since this strong assumption is often invalid in real-world scenarios, we choose to implicitly model the cross-modal correlation for large-scale multi-modal pretraining, which is the focus of the Chinese project ‘WenLan’ led by our team. Specifically, with the weak correlation assumption over image-text pairs, we propose a twotower pre-training model called BriVL within the crossmodal contrastive learning framework. Unlike OpenAI CLIP that adopts a simple contrastive learning method, we devise a more advanced algorithm by adapting the latest method MoCo into the cross-modal scenario. By building a large queue-based dictionary, our BriVL can incorporate more negative samples in limited GPU resources. We further construct a large Chinese multi-source imagetext dataset called RUC-CAS-WenLan for pre-training our BriVL model. Extensive experiments demonstrate that the pre-trained BriVL model outperforms both UNITER and OpenAI CLIP on various downstream tasks. Read More
#multi-modalMultimodal foundation models are better simulators of the human brain
Multimodal learning, especially large-scale multimodal pre-training, has developed rapidly over the past few years and led to the greatest advances in artificial intelligence (AI). Despite its effectiveness, understanding the underlying mechanism of multimodal pre-training models still remains a grand challenge. Revealing the explainability of such models is likely to enable breakthroughs of novel learning paradigms in the AI field. To this end, given the multimodal nature of the human brain, we propose to explore the explainability of multimodal learning models with the aid of non-invasive brain imaging technologies such as functional magnetic resonance imaging (fMRI). Concretely, we first present a newly-designed multimodal foundation model pre-trained on 15 million image-text pairs, which has shown strong multimodal understanding and generalization abilities in a variety of cognitive downstream tasks. Further, from the perspective of neural encoding (based on our foundation model), we find that both visual and lingual encoders trained multimodally are more brain-like compared with unimodal ones. Particularly, we identify a number of brain regions where multimodally-trained encoders demonstrate better neural encoding performance. This is consistent with the findings in existing studies on exploring brain multi-sensory integration. Therefore, we believe that multimodal foundation models are more suitable tools for neuroscientists to study the multimodal signal processing mechanisms in the human brain. Our findings also demonstrate the potential of multimodal foundation models as ideal computational simulators to promote both AI-for-brain and brain-for-AI research. Read More
#multi-modalMultimodality: A New Frontier in Cognitive AI
An exciting frontier in Cognitive AI involves building systems that can integrate multiple modalities and synthesize the meaning of language, images, video, audio and structured knowledge sources such as relation graphs. Adaptive applications like conversational AI; video and image search using language; autonomous robots and drones; and AI multimodal assistants will require systems that can interact with the world using all available modalities and respond appropriately within specific contexts. In this blog, we will introduce the concept of multimodal learning along with some of its main use cases, and discuss the progress made at Intel Labs towards creating of robust multimodal reasoning systems. Read More
Meta’s ‘data2vec’ is a step toward One Neural Network to Rule Them All
The race is on to create one neural network that can process multiple kinds of data — a more-general artificial intelligence that doesn’t discriminate about types of data but instead can crunch them all within the same basic structure.
The genre of multi-modality, as these neural networks are called, is seeing a flurry of activity in which different data, such as image, text, and speech audio, are passed through the same algorithm to produce a score on different tests such as image recognition, natural language understanding, or speech detection.
And these ambidextrous networks are racking up scores on benchmark tests of AI. The latest achievement is what’s called “data2vec,” developed by researchers at the AI division of Meta (parent of Facebook, Instagram, and WhatsApp).
The point, as Meta researcher Alexei Baevski, Wei-Ning Hsu, Qiantong Xu, Arun Babu, Jiatao Gu, and Michael Auli reveal in a blog post, is to approach something more like the general learning ability that the human mind seems to encompass. Read More
Rick's Cafe AI 