In human conversations, individuals can indicate relevant regions within a scene while addressing others. In turn, the other person can then respond by referring to specific regions if necessary. This natural referential ability in dialogue remains absent in current Multimodal Large Language Models (MLLMs). To fill this gap, this paper proposes an MLLM called Shikra, which can handle spatial coordinate inputs and outputs in natural language. Its architecture consists of a vision encoder, an alignment layer, and a LLM. It is designed to be straightforward and simple, without the need for extra vocabularies, position encoder, pre-/post-detection modules, or external plug-in models. All inputs and outputs are in natural language form. Referential dialogue is a superset of various vision-language (VL) tasks. Shikra can naturally handle location-related tasks like REC and PointQA, as well as conventional VL tasks such as Image Captioning and VQA. Experimental results showcase Shikra's prom

Amazon CEO Andy Jassy called generative A.I. “one of the biggest technical transformations of our lifetimes” in an interview with CNBC on Thursday. He also called many of today’s A.I. chatbots and other generative A.I. tools part of the “hype cycle,” declaring that Amazon was focused on the “substance cycle.”

Amazon’s bona fides in the space are well established, having been a player in artificial intelligence and machine learning long before the ChatGPTs and Bards of the world were publicly released. Former Fortune editor Brian Dumaine wrote a book in 2020 about how Amazon founder Jeff Bezos realized early on that imbuing machine learning into every facet of the company would allow it to gather data to constantly improve itself.

Much as it did with Amazon Web Services, which practically birthed the cloud computing industry that now powers the internet’s biggest companies, including its competitors, Amazon’s A.I. strategy is focused on cementing its position as a major player

For 3D object manipulation, methods that build an explicit 3D representation perform better than those relying only on camera images. But using explicit 3D representations like voxels comes at large computing cost, adversely affecting scalability. In this work, we propose RVT, a multi-view transformer for 3D manipulation that is both scalable and accurate. Some key features of RVT are an attention mechanism to aggregate information across views and re-rendering of the camera input from virtual views around the robot workspace. In simulations, we find that a single RVT model works well across 18 RLBench tasks with 249 task variations, achieving 26% higher relative success than the existing state-of-the-art method (PerAct). It also trains 36X faster than PerAct for achieving the same performance and achieves 2.3X the inference speed of PerAct. Further, RVT can perform a variety of manipulation tasks in the real world with just a few (∼10) demonstrations per task. Visual results, code, a

Covid-19 is said to cause long-term side effects in up to 67% of patients, and these health consequences can include chronic fatigue, loss of taste and smell and brain fog. Increasingly common too is Covid-related hair loss. Known as telogen effluvium, this phenomenon manifests as clumps of hair falling out after brushing or washing your hair.

It’s normal to shed hair daily – we lose about 100-150 hairs each day as hair drops from follicles to make way for new hair growth. This growth cycle occurs because 90% of the hair on our heads is in a growth phase (called anagen), while the remaining 10% is in a resting phase (called telogen). Anagen lasts for about three years before transitioning into the shorter telogen phase, following which hair is shed.

A stressful event like childbirth, certain medications, intense psychological stress and Covid-19 can trigger our bodies to shift a greater-than-normal proportion of growing anagen hairs into a resting telogen state, according to the

Recent work suggests that interpolating between the weights of two specialized language models can transfer knowledge between tasks in a way that multi-task learning cannot. However, very few have explored interpolation between more than two models, where each has a distinct knowledge base. In this paper, we introduce Derivative Free Weight-space Ensembling (DFWE), a new few-sample task transfer approach for open-domain dialogue. Our framework creates a set of diverse expert language models trained using a predefined set of source tasks. Next, we finetune each of the expert models on the target task, approaching the target task from several distinct knowledge bases. Finally, we linearly interpolate between the model weights using a gradient-free-optimization algorithm, to efficiently find a good interpolation weighting. We demonstrate the effectiveness of the method on FETA-Friends outperforming the standard pretrain-finetune approach.

The idea is simple - Specify what you want to research, and the AI will autonomously research it for you in minutes!

▸ One prompt generates an unbiased, factual and in depth research report

▸ Generate research, outlines, resource and lessons reports

▸ Aggregates over 20 web sources per research

▸ Includes an easy to use web interface

▸ Open source: https://github.com/assafelovic/gpt-researcher

▸ Scrapes web sources with javascript support

▸ Keeps track and context of visited and used web sources

Abstract:

Since the first laser was invented, the pursuit of high-energy lasers (HELs) has always been enthusiastic. The first revolution of HELs was pushed by the fusion of laser and aerospace in the 1960s, with the chemical rocket engines giving fresh impetus to the birth of gas flow and chemical lasers, which finally turned megawatt lasers from dream into reality. Nowadays, the development of HELs has entered the age of electricity as well as the rocket engines. The properties of current electric rocket engines are highly consistent with HELs’ goals, including electrical driving, effective heat dissipation, little medium consumption and extremely light weight and size, which inspired a second fusion of laser and aerospace and motivated the exploration for potential HELs. As an exploratory attempt, a new configuration of diode pumped metastable rare gas laser was demonstrated, with the gain generator resembling an electric rocket-engine for improved power scaling ability.

Original title: Focused Transformer: Contrastive Training for Context Scaling

Large language models have an exceptional capability to incorporate new information in a contextual manner. However, the full potential of such an approach is often restrained due to a limitation in the effective context length. One solution to this issue is to endow an attention layer with access to an external memory, which comprises of (key, value) pairs. Yet, as the number of documents increases, the proportion of relevant keys to irrelevant ones decreases, leading the model to focus more on the irrelevant keys. We identify a significant challenge, dubbed the distraction issue, where keys linked to different semantic values might overlap, making them hard to distinguish. To tackle this problem, we introduce the Focused Transformer (FoT), a technique that employs a training process inspired by contrastive learning. This novel approach enhances the structure of the (key, value) space, enabling an extensio

The digital plane we traverse is volatile and ever-changing, evolving at a pace that leaves the old world gasping for breath. One of the newest entrants in the tech theatre is the use of AI in content creation, specifically television. Fable, a San Fransisco-based start-up, is pioneering this bold frontier. Their brainchild, aptly named the Showrunner AI technology, or ‘SHOW-1’, aims to generate TV shows with viewers as starring roles.

I just copy/pasted what's in the link so formatting may be broken:

GPT-4's details are leaked.

It is over.

Everything is here: twitter.com/i/web/status/1… Parameters count:

GPT-4 is more than 10x the size of GPT-3. We believe it has a total of ~1.8 trillion parameters across 120 layers. Mixture Of Experts - Confirmed.

OpenAI was able to keep costs reasonable by utilizing a mixture of experts (MoE) model. They utilizes 16 experts within their model, each is about ~111B parameters for MLP. 2 of these experts are routed to per forward pass. MoE Routing:

While the literature talks a lot about advanced routing algorithms for choosing which experts to route each token to, OpenAI’s is allegedly quite simple, for the current GPT-4 model.

There roughly ~55B shared parameters for attention. Inference:

Each forward pass inference (generation of 1 token) only utilizes ~280B parameters and ~560 TFLOPs. This contrasts with the ~1.8 trillion parameters and ~3,700 TFLOP that

Abstract:

We observe that pre-trained large language models (LLMs) are capable of autoregressively completing complex token sequences -- from arbitrary ones procedurally generated by probabilistic context-free grammars (PCFG), to more rich spatial patterns found in the Abstract Reasoning Corpus (ARC), a general AI benchmark, prompted in the style of ASCII art. Surprisingly, pattern completion proficiency can be partially retained even when the sequences are expressed using tokens randomly sampled from the vocabulary. These results suggest that without any additional training, LLMs can serve as general sequence modelers, driven by in-context learning. In this work, we investigate how these zero-shot capabilities may be applied to problems in robotics -- from extrapolating sequences of numbers that represent states over time to complete simple motions, to least-to-most prompting of reward-conditioned trajectories that can discover and represent closed-loop policies (e.g., a stabilizi

cross-posted from: https://lemmy.world/post/1750098

Introducing Llama 2 - Meta's Next Generation Free Open-Source Artificially Intelligent Large Language Model

It's incredible it's already here! This is great news for everyone in free open-source artificial intelligence.

Llama 2 unleashes Meta's (previously) closed model (Llama) to become free open-source AI, accelerating access and development for large language models (LLMs).

This marks a significant step in machine learning and deep learning technologies. With this move, a widely supported LLM can become a viable choice for businesses, developers, and entrepreneurs to innovate our future using a model that the community has been eagerly awaiting since its initial leak earlier this year.