Tuesday, November 18, 2025
ad
Home Blog Page 133

Google AI Proposes ReAct: AI for Large Language Models

ReAct

Many organizations like Meta, OpenAI, and Google are working in the language domain and expanding language models to proximate more complex tasks. Language models generally transform and generate qualitative information, just like humans. Fundamentally, these models interpret data using algorithms that process information in the context of natural language. Once the algorithms work, the models accurately produce new content. Working in the same domain, Google AI has proposed a new artificial intelligence-driven approach, “ReAct,” for large language models. In this research, researchers have combined reasoning and acting advances to enhance the efficiency of language models. 

Existing language models usually work using two main techniques: chain-of-thought or pre-trained models. The models that work via chain-of-thought, a standard prompting method that enables a model to decompose the problem into numerous intermediate steps, are very efficient. With this prompting technique, language models with sufficient scale (approximately ~100B parameters) can also effectively solve reasoning problems. However, this technique makes reason-only models unsuitable for external environments and has limited exploring abilities. 

On the other hand, others that use pre-trained language models focus on mapping text contexts to actions with the model’s internal knowledge. These models are hence known as act-only models. However, even these models cannot reason or remain consistent in their actions as they learn from what has been fed. If the input is not socially sound and consistent, the model will learn from it and output in the same manner. Consequently, language models are known to exhibit more social bias compared to human-written text.

With ReAct, the researchers show that the Reason+Act (ReAct) paradigm outperforms models with reason-only and act-only paradigms. Especially when it is a large model, optimizing smaller models, and enhancing interpretability, ReAct is very efficient. To set up the ReAct prompting method, the PaLM-540B language model was used to prompt in-context domain-specific examples in the model. While executing reasoning-based tasks like navigating, the reasoning and acting jobs are alternated. For instance, say the prompt is “go to” in a room for navigating, then this command needs a task-solution trajectory that comprises multiple reasoning-action-observation stages. 

Read More: Meta Turned Down the Galactica Demo After Being Criticized as “Dangerous”

What sets the ReAct approach apart is that the reasoning traces only need to be sparsely located throughout the trajectory of tasks with a large number of actions. The ReAct model then determines when and how reasoning and action responses will occur asynchronously. The PaLM-540B model was used to generate successful trajectories, which were later used to fine-tune smaller models like PaLM-8/62B.

The researchers evaluated ReAct against four benchmarks to see if it could reduce the extensive need for human annotation. These benchmarks were: HotPotQA (for question answering), Fever (for fact-checking), ALFWorld (text-based gaming), and WebShop (web page navigation. In the context of HotPotQA and Fever, it was observed that the model overcomes standard errors and hallucinations in chain-of-thought reasoning. When it comes to ALFWorld and WebShop, ReAct surpasses reinforcement learning techniques.

ReAct was also investigated using a human inspector control over its reasoning traces so that the researchers could evaluate human interactions with the model. ReAct successfully altered its behavior corresponding to the revisions provided by the human inspector. The model generates a hallucinatory trajectory when any improvised revision is entered, making it highly efficient in human-machine interaction with negligible human involvement.
Google has been actively working on language models, and this new model is yet another stride of success in that direction. As seen in the paper, ReAct makes it feasible to describe a behavior or feedback within a model while it flexibly handles the input and calls for action. Be it multiple-choice question-answering, fact-verification, or interactive decision-making, ReAct exhibits a commendable performance.

Advertisement

Meta Turned Down the Galactica Demo After Being Criticized as “Dangerous”

meta galactica criticized as dangerous

Meta introduced a new large language model ‘Galactica’ to generate original academic papers with simple prompts. But as quickly as it was introduced, several people criticized it as “dangerous,” after which Meta turned down the demo. 

Earlier, on visiting the website, users could see an option to “Generate” content, as seen in the image below.

However, as more and more people reported it to be full of “statistical nonsense” and that it was generating “wrong” content, the website withdrew the option for people to experiment with. 

Grady Booch, a software engineer described Galactica as “little more than statistical nonsense at scale.” He said it was amusing but IMHO (International Human Rights & Media Organization) unethical.

Gary Marcus tweeted his concern after reviewing Galactica and said that it has “jumped the AI shark.” 

He also added that Galactica “prevaricates” a lot, implying that it is evasive when it comes to mentioning the exact truth. He proceeded to say that students would love to use such a model to intimidate their teachers, while others, aware of the risks, should be terrified. 

Michael Black, director of the Max Planck Institute for Intelligent Systems, tweeted:

He said the work is an interesting advancement, yet it is not useful and safe for doing scientific work. He used the word “dangerous” and explained that Galactica outputs grammatically coherent content, but there is no certainty of it being unbiased and scientifically correct. In such ambiguity, if these results slip into scientific submissions, it would be potentially distorting. 

Read More: Meta Will Host a ‘Hyperrealistic’ Avatar of Late Rapper Biggie in a VR Concert

He feared that models like Galactica could usher in an era of scientific deep fakes. He said, “Alldieck and Pumarola will get citations for papers they didn’t write. These papers will then be cited by others in real papers. What a mess this will be.”

Associate Professor Keenan Crane, Computer Science, and Robotics at CMU, also expressed distrust in Galactica. He said that none of the deep language models could be trusted completely as they are intuitive and authoritative and imitate reliability. 

Advertisement

Meta India appoints Sandhya Devanathan as its new Vice President 

Meta India Sandhya Devanathan Vice President

Meta announced on Thursday it has appointed Sandhya Devanathan as the new vice president of its India unit. She will be the head of the company’s business in the country.

Meta said Devanathan’s journey as the Vice President of Meta India would begin on January 1, and she will relocate to India to head the organization. She is currently based in Singapore.

Devanathan is succeeding Ajit Mohan, who resigned on November 3. Meta India’s director and head of partnerships, Manish Chopra, has taken over Mohan’s position on an interim basis.

Read More: Meta Will Host A ‘Hyperrealistic’ Avatar Of Late Rapper Biggie In A VR Concert

She joined Meta in January 2016 as the Group Director of SEA – Travel, FinServ, and E-commerce, in Singapore. She became the tech giant’s Business Head for Vietnam and Managing Director for Singapore in August. She is also Meta’s Vice President for Gaming in the Asia-Pacific region which she assumed in April 2020.

Devanathan also serves, or has served, as a board member of several organizations. These include the National Library Board (Singapore), Singapore Management University, Pepper Financial Services Group, Women’s Forum for the Economy and Society, and Ministry of Information and Communications (Singapore).

Advertisement

Motional and Lyft team up to Launch Robotaxi Service in Los Angeles

Motional Lyft robotaxi service los angeles

Motional, the joint venture between Hyundai and Aptiv for autonomous vehicles, is bringing its robotaxis to Los Angeles, where riders can request them using the Lyft app. The service is provided by Motional’s fleet of electric Hyundai Ioniq 5 cars, all of which will be operated entirely autonomously upon service launch and without the need for a human safety driver. However, neither party has stated when the service would be available.

After Las Vegas, Los Angeles is the second location in Lyft and Motional’s multi-city collaboration, where the two businesses are offering shared trips using Motional’s IONIQ 5 AVs. The IONIQ 5 cars, like those in Las Vegas, will be incorporated into Lyft’s Los Angeles network. Riders in Los Angeles will have access to a rideshare network with various transportation options, including autonomous cars and conventional ridesharing, to get them where they need to go. People can use the Lyft app to open the doors when the car arrives, and there is a passenger display in every car that can be used to contact a remote agent anytime.

The IONIQ 5 is powered by Motional’s AI-first autonomous vehicle stack, which includes over 30 sensors, including cameras, radar, and LIDAR, to reliably identify objects at extremely long distances. An onboard ‘compute’ system analyzes all the data coming in from the robotaxi cameras and sensors.

Motional claims to have used Lyft to provide over 100,000 trips in Las Vegas, with a five-star rating rate of more than 95%. Additionally, the company signed agreements to a 10-year partnership with Uber, announcing that it would begin providing passenger trips later this year and that its vehicles will be “strategically placed” in US cities.

Read More: Israel to Trial Self-driving bus Pilot Program worth US$17 million

Since 2016, Motional has established an office in Los Angeles. Last year, it announced plans to expand its foothold in California by establishing a new operations center in Los Angeles to facilitate testing on public roads, recruiting additional engineers, and creating an office in Silicon Valley. However, this will be the first time LA citizens can use the Lyft app to book an autonomous vehicle. This implies that Motional must get all necessary permits before accepting passengers. AV companies must get a series of permits from the Department of Motor Vehicles and Public Utilities Commission to legitimately shuttle passengers and receive ride payments.

Despite the mandatory permissions, California has been one of the most supportive and legally friendly states when it comes to robotaxis and fully autonomous vehicles. A month ago, the Waymo unit of Alphabet revealed intentions to launch a self-driving robotaxi service in key Los Angeles districts. In September, UberEats announced the introduction of Nuro’s driverless delivery robots in California.

Advertisement

Meta Will Host a ‘Hyperrealistic’ Avatar of Late Rapper Biggie in a VR Concert

meta will host late biggie avatar in a VR concert

Earlier this year, Meta revealed its plans to host a hyperrealistic avatar of The Notorious B.I.G, a late rapper popularly known as “Biggie,” in a VR concert called The Brook. The concert was meant to accompany other upcoming NFTs that were due to launch by the end of 2022. 

Now, Meta has updated the story as it announced that its Horizon Worlds will host ‘The Notorious B.I.G. Sky’s The Limit: A VR Concert Experience.’ As per the announcement, there will be a ‘hyperrealistic’ avatar of Biggie, with legs and a human-like face. 

Biggie, or Christopher Wallace, was a well-known hip-hop rapper who rose to stardom in the early 90s. He was mysteriously killed in a drive-by shooting in March 1997. 

Read More: Meta Introduces ‘Galactica,’ an AI System that Generates Academic Papers from Simple Text Inputs

The VR concert will be organized in collaboration with the late icon’s estate managers and will take place on December 16, 2022. The rapper’s avatar will be seen performing his famous raps, just like the ‘Hypnotize.’ He will also take concert attendees through a “narrative journey” of Biggie’s life in Brooklyn. 


The iconic rapper has received such homage before. On what would have been his 50th birth anniversary, May 2021, the Empire State Building changed its color to red and white, with a spinning crown. A special memorial, attended by people like Lil Kim, Lil’ Cease, and Voletta Wallace (his mother), was also held in New York.

Advertisement

COP27: Tuvalu to become first Metaverse Nation. Is this answer to Climate Change?

Tuvalu metaverse COP27

The island nation of Tuvalu claims it will have no alternative but to create a digital version of itself since the temperature of the planet is projected to increase by as much as 2.8°C by the end of the century. In a disheartening speech to world leaders at COP27, the small, low-lying Pacific Island nation confirmed that it has been forced to become “the world’s first digital nation” due to inaction on climate change. Tuvalu’s foreign minister, Simon Kofe, said at the COP27 climate meeting on Tuesday that the island nation will resort to the metaverse to preserve its culture and heritage due to rising sea levels.

Kofe made headlines at the COP26 meeting last year when he publicly gave a speech while knee-deep in the ocean to emphasize the grave threat that climate change posed to island nations like Tuvalu. In his most recent video speech, the idea that all nations could end up living in the metaverse is put up as a possible outcome.

The world has eight years to cut yearly global emissions by a further 45% compared to projections based on existing policies in order to meet the 1.5°C target outlined by the Paris Agreement and prevent visibly worse impacts of global warming. An additional 30% decrease in emissions is required to keep the temperature increase to within 2°C.

Kofe said, “Our land, our ocean, our culture are the most precious assets of our people — and to keep them safe from harm, no matter what happens in the physical world, we’ll move them to the cloud.”

Tuvalu is the fourth-smallest country in the world by area, with a total land area of only 10 square miles, divided among its nine low-lying islands. With about 12,000 residents, it is also the third-smallest city by population. During high tide, approximately 40% of Tuvalu’s capital district gets submerged. The first step in the digitization project will be to recreate Teafualiku Islet, the smallest island in Tuvalu and the first area of the nation that would be lost if sea levels keep rising.

Considering the small landmass size of Tuvalu, it will be a lot less difficult to create a digital twin of the nation compared to bigger nations like Australia and China. However, given the level of authenticity, this project would undoubtedly aim for, digitally reproducing a nation will still be a laborious and technologically challenging task. In general, so far, metaverse visuals aren’t exactly impressive either, so there is a disappointing possibility that digital Tuvalu will ultimately be lost along the way.

Read More: Data Center and Climate Change Tango: How do they impact each other?

Mr. Kofe urged citizens all across the world to take action by holding their governments accountable for their Paris Agreement obligations. Tuvalu has even created a website with pre-written letters of action that citizens could use to voice their concerns to officials.

In addition to making a physical transition online, the government is considering ways to digitize its administrative system and move its activities online so that it can continue to effectively perform its responsibilities as a sovereign state in the digital era. Additionally, it is investigating legislative options to guarantee the permanence of its sovereignty and marine borders despite rising sea levels.

Kofe suggests that the following three facets of Tuvaluan identity might be replicated in the metaverse:

  1. Territory: the reproduction of Tuvalu’s natural beauty, which can be explored in a variety of ways. 
  2. Culture: Tuvaluans’ shared language, traditions, and practices must be maintained in order for them to connect with one another, wherever they may be.
  3. Sovereignty: Could the government of Tuvalu have authority over virtual land in its place if it were to lose control of the terrestrial land it now controls. 

According to Kofe, seven states have consented to continue recognition, but there might be problems if Tuvalu fails because it is a relatively new area of international law.

Tuvalu might be following the footsteps of nations that have already announced metaverse plans. Barbados, an island nation in the Caribbean, became the first country to establish an embassy in the Decentraland metaverse in 2021. Even Seoul had announced creating Metaverse Seoul – a metaverse ecosystem for all administrative functions, including the economy, culture, tourism, education, and civil grievances. Earlier this year, an Australian indigenous group also announced plans to establish an embassy in the metaverse.

It is important to remember that Kofe could be aware that the metaverse is not Tuvalu’s solution to its troubles. He makes it clear that we must concentrate on lessening the effects of climate change through efforts like a convention on the non-proliferation of fossil fuels. His provocational video speech about Tuvalu entering the metaverse is bait to achieve resounding success in getting media attention. Even naively, it is foolish to think that migrating to the metaverse is a feasible remedy to climate change. 

This is because the virtual world of metaverse is powered by servers, data centers, VR headsets, and other devices, are connected via the internet – implying they are energy hungry and have a massive carbon footprint. Thus, considering resource-intensive metaverse as an answer to climate change can have a misleading precedent.

Metaverse can serve as an institution that preserves the cultural heritage and territorial identity of a place but not as a sovereign state, at least for now. The real world is not always a laboratory for fictitious blueprint concepts.

Advertisement

NVIDIA collaborates with Microsoft to build massive AI supercomputer

On 16th November, NVIDIA announced a multi-year collaboration with Microsoft to develop one of the best powerful supercomputers in the world by using Microsoft Azure’s advanced supercomputing infrastructure with NVIDIA GPUs, networking, and the entire stack of AI software.

Azure’s cloud-based AI supercomputer consists of robust and scalable ND- and NC- series virtual machines optimized for AI-distributed training and inference. Microsft is the first public cloud to include NVIDIA’s advanced stack and ten thousand NVIDIA A100 and H100 GPUs. It includes NVIDIA Quantum-2 400 Gb/s InfiniBand networking and the NVIDIA AI enterprise software suite.

Read more: Amazon introduces its new item-picking robot: Sparrow

 The vice president of enterprise computing at NVIDIA, Manuvir Das, mentioned that the development of foundation models had given rise to a tidal wave of research, encouraging startups and enabling new enterprise applications. The NVIDIA and Mircosoft collaboration can offer researchers and companies state-of-the-art AI infrastructure to capitalize on the transformative power of AI.

As per NVIDIA, after the successful implementation of the supercomputer, customers would deploy thousands of GPUs on a single cluster to build complex AI systems. However, NVIDIA should have disclosed details on when the new supercomputer will be ready. Since they have mentioned a ‘multi-year’ collaboration, it can take time to develop.

Advertisement

Rajasthan School Education Department has Set a World Record by Assessing 1.35cr OCR Sheets with AI

education department set world record assessing sheets with ai

The Rajasthan School Education Department has set a world record by assessing over 1.35cr OCR sheets with AI in a state-level mass drive. The department’s ‘Rajasthan Ke Shiksha Main Badhte Kadam’ (RKSMBK) application assessed the sheets belonging to students of classes 3 to 8. The AI assessment record has made a place in the World Book of Records, London.

Baluki Das Kalla, Primary and Secondary Education Minister, wrote in a social media post that Aditi Tank and Pratham Bhalla, World Book of Records representatives, presented the certificate to the department.

Read More: Amazon introduces its new item-picking robot: Sparrow.

The RKSMBK application is a part of the ‘Steps in Education of Rajasthan’ program. The government of Rajasthan is working to develop infrastructure in the education sector for primary, secondary, and higher educational institutions. 


The initiatives are focused on enhancing the employability quotient of the state by bridging the gap between education and industry requirements. This is one of many technology-driven programs undertaken by the state government. There are many other initiatives like the SMILE (Social Media Interface for Learning Engagement) program, Shikshadarshan (Educational content via TV), Kala Utsav summer camps, E-content invited via ShalaDarpan and many others.

Advertisement

IIT Kharagpur announces free online course in artificial intelligence and machine learning

IIT Kharagpur free AI course

IIT Kharagpur announced a free online course called applied linear algebra in artificial intelligence and machine learning that can be completed in 12 weeks. 

The course is being offered on the SWAYAM NPTEL platform and would be most suited for students with knowledge about engineering mathematics, along with some exposure to linear algebra. Participants will also receive a certificate provided they complete the course and take an examination by paying Rs 1000. 

The course will be conducted from 23 January 2023 to 14 April 2023, and the last date to enroll for the course is 23 January 2023.

Read More: Amazon Introduces Its New Item-Picking Robot: Sparrow

The course will be conducted by professor Swanand Khare who obtained M.Sc. and Ph.D. degrees from IIT Bombay in 2005 and 2011, respectively. He is working as an associate professor in the mathematics department and the center of excellence in artificial intelligence at IIT Kharagpur. 

The course would be most suitable for senior undergraduate and postgraduate students from streams such as computer science engineering, electrical engineering, artificial intelligence, electronics and communication engineering, and mathematics.

Advertisement

Notion AI tests new feature to write blog posts, email templates

Notion AI

Notion is starting to test a new feature called Notion AI, which according to Notion CEO Ivan Zhao, could help users to write a blog post, a list of great business books to read, a recruiting email template, etc. All users have to do is tell the app what they want, and the app creates the content right before their eyes.

According to Zhao, the first thing Notion AI is writing. Work requires a lot of writing, some of it repetitive, like creating job descriptions, and some creative, like brainstorming new ideas or writing blog posts. Zhao thinks Notion AI can help with both.

“Notion solves the cold start problem for many users in this space. And even more than that, it saves time. Notion AI will not write a publishable blog post, but it can help skip the first draft phase and jump straight into editing,” said Zhao.

Read More: Amazon Introduces Its New Item-Picking Robot: Sparrow

Most of the data in Notion AI’s models comes from the broader internet. According to Zhao, the company is partnering with AI providers. It plans to personalize the model to companies so that the AI can write in the voice of the company and use its internal information.  

Notion is still figuring out what Notion AI can be best suited for and how to help users get familiar with how to use the new tech. They opened the Notion AI waitlist yesterday, with plans to open Notion AI to product ambassadors first, then more users over time. 

Advertisement