Chain-of-Thought (CoT) prompting can effectively elicit complex multi-step reasoning from Large Language Models~(LLMs). For example, by simply adding CoT instruction ``Let's think step-by-step'' to each input query of MultiArith dataset, GPT-3's accuracy can be improved from 17.7\% to 78.7\%. However, it is not clear whether CoT is still effective on more recent instruction finetuned (IFT) LLMs such as ChatGPT. Surprisingly, on ChatGPT, CoT is no longer effective for certain tasks such as arithmetic reasoning while still keeping effective on other reasoning tasks. Moreover, on the former tasks, ChatGPT usually achieves the best performance and can generate CoT even without being instructed to do so. Hence, it is plausible that ChatGPT has already been trained on these tasks with CoT and thus memorized the instruction so it implicitly follows such an instruction when applied to the same queries, even without CoT. Our analysis reflects a potential risk of overfitting/bias toward instructions introduced in IFT, which becomes more common in training LLMs. In addition, it indicates possible leakage of the pretraining recipe, e.g., one can verify whether a dataset and instruction were used in training ChatGPT. Our experiments report new baseline results of ChatGPT on a variety of reasoning tasks and shed novel insights into LLM's profiling, instruction memorization, and pretraining dataset leakage.

The proliferation of Large Language Models (LLMs), such as ChatGPT, has raised concerns about their potential impact on academic integrity, prompting the need for LLM-resistant exam designs. This article investigates the performance of LLMs on exams and their implications for assessment, focusing on ChatGPT's abilities and limitations. We propose guidelines for creating LLM-resistant exams, including content moderation, deliberate inaccuracies, real-world scenarios beyond the model's knowledge base, effective distractor options, evaluating soft skills, and incorporating non-textual information. The article also highlights the significance of adapting assessments to modern tools and promoting essential skills development in students. By adopting these strategies, educators can maintain academic integrity while ensuring that assessments accurately reflect contemporary professional settings and address the challenges and opportunities posed by artificial intelligence in education.

According to the NYT, Google's position is so threatened that Samsung is considering replacing Google with Bing as the default search engine on its mobile devices. This deal is worth an estimated $3 billion in annual revenue to Google (the company has a similar deal with Apple worth roughly $20 billion), though it's not clear how seriously Samsung is considering the switch. The company may have been been swayed by Microsoft's AI work, but it might also be simply taking advantage of Google's moment of weakness.

This project is under construction and we will have all the code ready soon. Language serves as an interface for LLMs to connect numerous AI models for solving complicated AI tasks! We introduce a collaborative system that consists of an LLM as the controller and numerous expert models as collaborative executors (from HuggingFace Hub). However, it means that Jarvis is restricted to models running stably on HuggingFace Inference Endpoints. Now you can access Jarvis' services by the Web API.

Russell Wald, director of the Stanford Institute for Human-Centered AI, sounds off on'The Story.' After a Stanford survey of researchers showed more than a third believe that AI (artificial intelligence) could cause a "nuclear-level catastrophe," one of the directors of the California college's Institute for Human-Centered Artificial Intelligence said another key to that fear is deciphering what values system the human-inputted but self-determining bots will hold. "I think that there is a little bit of overwhelming concern by a minority of people who are developing this technology. And one part of that is the concern that this technology will spin out of control and out of our hands," said Russell Wald, the institute's managing director for policy and society, said Monday on "The Story with Martha MacCallum." "I think that is a probably fairly limited subset. But the bigger issue here I think we need to look at is who has a seat at this table -- and ensuring that we have a more diverse set of people at the table so that we can get out of this hysteria a little bit."

Sooner than even the most pessimistic among us have expected, a new, evil artificial intelligence bent on destroying humankind has arrived. Known as Chaos-GPT, the autonomous implementation of ChatGPT is being touted as "empowering GPT with Internet and Memory to Destroy Humanity." It hasn't gotten very far. But it's definitely a weird idea, as well as the latest peculiar use of Auto-GPT, an open-source program that allows ChatGPT to be used autonomously to carry out tasks imposed by the user. AutoGPT searches the internet, accesses an internal memory bank to analyze tasks and information, connects with other APIs, and much more--all without needing a human to intervene.

As the contract lifecycle management company Ironclad is today releasing its AI redlining tool AI Assist out of beta, is has revealed that the tool is powered by OpenAI's GPT-4, making it what Ironclad says is the first contract redlining application powered by the latest version of Open AI's generative AI. "The results with AI Assist have been beyond what we could even have imagined," said Ironclad CEO and co-founder, Jason Boehmig. "An initial pass at contract redlining usually takes about 40 minutes. Already, some large enterprises are using Ironclad AI to review over 50% of their incoming contracts, so the compounding business impact there is unprecedented." Although Ironclad says that this is the first redlining tool to use GPT-4, Casetext's CoCounsel, which is built on GPT-4, has capabilities for checking contract policy compliance and suggesting redlines to bring contracts into compliance. It should also be noted that there are other contract redlining tools on the market that use AI, but not GPT-4.

Google and it's very own version of Chatbot Bard is soon to be released to the public. Recently Google AI engineers stated that it's opening up access to Bard, the Google's very own AI-powered chatbot that's a rival to services released by competitor Microsoft and it's OpenAI. Google is starting with users in the US and UK, who can go to the Bard site to sign up for the waiting list. "We've learned a lot so far by testing Bard, and the next critical step in improving it is to get feedback from more people," stated Google's Sissie Hsiao and Eli Collins. Google has announced Bard and the company went into "code red" following the release of the OpenAI's ChatGPT late last year to garner it's own AI Chat.

Fox News Flash top headlines are here. Check out what's clicking on Foxnews.com. An education software company has developed a program it says schools and universities can use to detect whether students are using AI to complete their tests and essays, according to a new report. The company, Turn It In, has a long history of developing tools educators can use to detect plagiarism. The company has now turned to an AI system that it says can effectively determine whether students are responsible for their own work, or whether they turned to an AI like ChatGPT. Turn It In's tool isn't foolproof, however, according to a test conducted at the University of Southern California.

Elon Musk Creates new AI Company to rival OpenAI as Elon Musk announced his plans to enter in the Artificial Intelligence (AI) market of worth more than US$ 125 billion, globally. Since, the global AI market is expected to hit US$ 1,591 billion by 2030 with a registered CAGR of 38.1% from 2022 to 2030, Elon Musk has planned to enter this arena. The boss of world's leading companies like Twitter, Telsa, StarLink, and SpaceX has announced a new venture and as he has registered a company called'X.AI'. It is reveled that the new subsidiary will be the home of efforts to build new AI based tools similar to ChatGPT that is currently owned by OpenAI. To accelerate with X.AI, Musk is reported that he is assembling a team of AI experts and researchers as he is discussing with with investors of SpaceX and Tesla about putting money into this new venture.