What is the fastest way to estimate the value of a car? How to do it with only one photo? How to do it for free? An open-source, multi-modal, zero-shot model solves this problem in 3 sec. What are the unexplored use cases of Large Language Models like ChatGPT?

SINGAPORE - Teachers in Singapore say they will likely have to move from assignments requiring regurgitation to those that require greater critical thinking, to stay ahead in the fight against plagiarism. This comes on the back of the rise of ChatGPT, an intelligent chatbot that is able to spin essays and solve mathematical equations in seconds. ChatGPT, developed by San Francisco research firm OpenAI, is being talked about as a major step forward in artificial intelligence (AI), especially its latest version released in November. Simple and free of charge – for now – the chatbot has prompted some schools to start thinking of ways to mitigate cheating, ahead of students' return to schools. ChatGPT's ability to break down complicated concepts into simple language and respond to follow-up questions logically has raised concerns over whether existing plagiarism detection software used in schools, such as Turnitin, can sniff out text drafted by bots.

This year, we see significant progress in the field of generative models. Stable Diffusion creates hyperrealistic art. ChatGPT answers questions to the meaning of life. Galactica learns humanity's scientific knowledge but also reveals the limitations of large language models. This article is my take on the 20 most impactful AI papers of 2022.

Recent ChatGPT has managed to gather the attention from people of varying fields as it amazes with its conversational capabilities. Among other things, it can produce code and resolve problems at a blink of an eye making life easier for a lot of people. "It solves problems at a blink of eye and knows programming" you said. That means… content creators, programmers and data scientists, to name a few, are doomed! If that's your first thought when you heard about ChatGPT's capabilities, I won't blame you, its what probably everyone thought anyways, at first at least.

OpenAI's ChatGPT introduced a way to automatically create content but plans to introduce a watermarking feature to make it easy to detect are making some people nervous. This is how ChatGPT watermarking works and why there may be a way to defeat it. ChatGPT is an incredible tool that online publishers, affiliates and SEOs simultaneously love and dread. Some marketers love it because they're discovering new ways to use it to generate content briefs, outlines and complex articles. Online publishers are afraid of the prospect of AI content flooding the search results, supplanting expert articles written by humans.

Closed book Abstractive: These systems do not have access to external data store. They store information internally in the model parameters. ChatGPT and other large language models are part of this category. Unlike open book systems, these system do not have access to the latest information.

We achieved 147 TFLOP/s/GPU utilization on NVIDIA's 80 GB A100 GPUs, roughly 17 percent higher than published by NVIDIA researchers on similar hardware. Expect to see Meta on this list again. They have adopted a completely open-source approach, where they share their models, training data, logs, and a lot more. This is an unprecedented move with a lot of implications for the Machine Learning sector. To learn more about this (and the OPT Model), check out the following article. The OPT Model has a lot of potential when it comes to being a GPT replacement, given that it seems to be designed as an alternative to it. However, Meta is not the only Tech Giant with a horse in the race.

The expectation is mounting up around OpenAI's ChatGPT-4, which is scheduled for 2023, although there is no official confirmation on either the launch or beta testing of it. GPT-4 stands for Generative Pre-trained Transformer 4. It's basically an artificial intelligence system that can create human-like text. While the current ChatGPT-3 has 175 billion parameters, ChatGPT-4 might have 1 trillion, or even more, according to some reports. Similarly, it will be capable of text answering, content generation, language translation, and text summarization, just like the current ChatGPT-3. The increase of parameters -- a measure of the complexity of the neural machine to do useful things -- should enable ChatGPT-4 to produce more accurate responses at a much faster rate.

In this paper, hypernetworks are trained to generate behaviors across a range of unseen task conditions, via a novel TD-based training objective and data from a set of near-optimal RL solutions for training tasks. This work relates to meta RL, contextual RL, and transfer learning, with a particular focus on zero-shot performance at test time, enabled by knowledge of the task parameters (also known as context). Our technical approach is based upon viewing each RL algorithm as a mapping from the MDP specifics to the near-optimal value function and policy and seek to approximate it with a hypernetwork that can generate near-optimal value functions and policies, given the parameters of the MDP. We show that, under certain conditions, this mapping can be considered as a supervised learning problem. We empirically evaluate the effectiveness of our method for zero-shot transfer to new reward and transition dynamics on a series of continuous control tasks from DeepMind Control Suite. Our method demonstrates significant improvements over baselines from multitask and meta RL approaches.

Large pretrained language models have recently conquered the area of natural language processing. As an alternative to predominant masked language modelling introduced in BERT, the T5 model has introduced a more general training objective, namely sequence to sequence transformation, which includes masked language model but more naturally fits text generation tasks such as machine translation, summarization, question answering, text simplification, dialogue systems, etc. The monolingual variants of T5 models have been limited to well-resourced languages, while the massively multilingual T5 model supports 101 languages. In contrast, we trained two different sized T5-type sequence to sequence models for morphologically rich Slovene language with much less resources and analyzed their behavior on 11 tasks. Concerning classification tasks, the SloT5 models mostly lag behind the monolingual Slovene SloBERTa model but are useful for the generative tasks.