OpenAl and Deepmind Employees Warn of Al Dangers Including Human Extinction, that Companies Are Hiding

There has been a significant & serious development regarding AI safety concerns. A group of current and former employees from OpenAI and Google's DeepMind have come forward with an open letter —righttowarn.ai, warning about the potential dangers associated with advanced AI technologies, including human extinction. They allege that these companies are prioritizing financial gains over safety and are not being transparent about the risks involved.

The letter emphasizes the need for better oversight and regulation to prevent serious harms, such as the further entrenchment of existing inequalities, manipulation, misinformation, and even the loss of control over autonomous AI systems. The employees are advocating for a culture of open criticism and are calling for solid whistleblower protections to enable the discussion of these risks without fear of retaliation.

 

This is a developing story, and it highlights the importance of ethical considerations and transparency in the field of AI development. It's crucial for AI companies to engage with governments, civil society, and other stakeholders to ensure that AI technologies are developed responsibly and safely.

Specific Risks Employees Are concerned

The employees from OpenAI and Google DeepMind have raised concerns about several specific risks associated with the development and deployment of advanced AI systems. These include:

Entrenchment of Existing Inequalities: Advanced AI could exacerbate social and economic disparities if its benefits are not distributed equitably.

Manipulation and Misinformation: AI systems could be used to create and spread false information, potentially influencing public opinion and undermining trust in institutions.

Loss of Control: There is a risk that autonomous AI systems could become uncontrollable, leading to unintended consequences.

Human Extinction: The letter mentions the extreme risk that unregulated AI poses, including scenarios that could lead to human extinction.

The group behind the open letter has urged AI firms to facilitate a process for current and former employees to raise risk-related concerns and not enforce confidentiality agreements that prohibit criticism. They emphasize the need for transparency and oversight to ensure that AI development does not compromise safety or ethical standards.

For the 1st Time, AI Beats Conventional Weather Forecasting by Accurately Predicting Weather 3-10 Days Ahead, in Less Than A Minute

For the first time, Artificial Intelligence (AI) models are leading in making weather predictions 3 to 10 days ahead.

The GraphCast Al model, developed by Google DeepMind, uses a machine-learning model that has learned from more than 40 years of weather forecasts.

It outperformed the conventional forecasting method in 90% of the 1,380 metrics used, which included temperature, pressure, wind speed and direction, and humidity at different levels of the atmosphere, Google DeepMind said in a peer-reviewed paper.

Developed by Google’s AI company DeepMind in London, GraphCast outperforms conventional and AI-based approaches at most global weather-forecasting tasks. Researchers first trained the model using estimates of past global weather made from 1979 to 2017 by physical models. This allowed GraphCast to learn links between weather variables such as air pressure, wind, temperature and humidity.

The trained model uses the ‘current’ state of global weather and weather estimates from 6 hours earlier to predict the weather 6 hours ahead. Earlier predictions are fed back into the model, enabling it to make estimates further into the future.

The standard conventional method called numerical weather prediction (NWP) uses mathematical models based on physical principles. These physical models crunch weather data from buoys , satellites and weather stations worldwide using supercomputers. The calculations accurately map out how heat, air and water vapour move through the atmosphere, but they are expensive and energy-intensive to run.

DeepMind researchers found that GraphCast could use global weather estimates from 2018 to make forecasts up to 10 days ahead in less than a minute, and the predictions were more accurate than the European Centre for Medium-Range Weather Forecasts (ECMWF)'s High RESolution forecasting system (HRES) — one version of the UK's NWP — which takes hours to forecast. Notably, ECMWF provides world-leading weather predictions up to 15 days in advance.

The GraphCast model can run from a desktop computer and makes more accurate predictions than conventional models in minutes rather than hours.

“GraphCast currently is leading the race amongst the AI models,” says computer scientist Aditya Grover at University of California, Los Angeles.

The model is described in Science on 14 November.

In the troposphere, which is the part of the atmosphere closest to the surface that affects us all the most, GraphCast outperforms HRES on more than 99% of the 12,00 measurements done by Deepmind researchers.

Across all levels of the atmosphere, GraphCast model outperformed HRES on 90% of weather predictions.

Earlier in last month, IndianWeb2.com reported that researchers from the Universities of California at Berkeley and Santa Cruz, and the Technical University of Munich, unveiled the Recurrent Earthquake foreCAST (RECAST), a deep learning model for improved earthquake forecasting.

In A World’s 1st in Drug Discovery, AI Develops A Treatment for Cancer in 30 Days

Researchers at the University of Toronto have developed a potential treatment for liver cancer using artificial intelligence

The team applied AlphaFold, an Al-powered protein structure database, to uncover a novel treatment pathway for cancer. The creation of the potential drug was accomplished in just 30 days, and the Al system can also predict a patient's survival rate.

According to the study published in the journal Chemical Science, researchers at the University of Toronto along with Insilico Medicine developed a potential treatment for hepatocellular carcinoma (HCC) — the most common type of primary liver cancer — with an AI drug discovery platform called Pharma.AI, an end-to-end AI-powered drug discovery platform.

The study is being touted as the world’s first to apply a groundbreaking AI technology, AlphaFold, to drug discovery research.

The researchers included a biocomputational engine, PandaOmics, and a generative chemistry engine, Chemistry42. They discovered a novel target for HCC – a previously undiscovered treatment pathway – and developed a “novel hit molecule” that could bind to that target without the aid of an experimentally determined structure.

The scientists first used AI to scan HCC for protein “weak spots.” Once one was detected, another program designed small molecules that could target and take down the specific protein, the paper says. Researchers then tested these molecules on live cells, one of which appeared effective at slowing cancer growth.

This feat was accomplished in just 30 days from target selection and after only synthesizing seven compounds.

“This paper is further evidence of the capacity for AI to transform the drug discovery process with enhanced speed, efficiency, and accuracy,” Michael Levitt, a Nobel Prize winner in chemistry, said. “Bringing together the predictive power of AlphaFold and the target and drug-design power of Insilico Medicine’s Pharma.AI platform, it’s possible to imagine that we’re on the cusp of a new era of AI-powered drug discovery.”

In a statement, Alex Zhavoronkov, founder and CEO of Insilico medicine, said, "While the world was fascinated with advances in generative AI in art and language, our generative AI algorithms managed to design potent inhibitors of a target with an AlphaFold-derived structure."

Insilico Medicine is a biotechnology company based in Hong Kong and New York.

AlphaFold is an artificial intelligence program developed by DeepMind, a subsidiary of Alphabet, the parent company of Google.

Last year, AlphaFold predicted protein structures for the whole human genome – a remarkable breakthrough in both AI applications and structural biology.

Google’s Parent Alphabet Forms New AI and DeepMind-powered Drug Discovery Company

Google’s parent company Alphabet Inc. has launched a new company --  Isomorphic Labs -- a commercial venture with the mission to reimagine the entire drug discovery process from first principles with an AI-first approach.

Isomorphic Labs is an artificial intelligence company built on the protein-folding simulation successes achieved at DeepMind, an AI subsidiary of Alphabet Inc. that was acquired by Google in 2014.

Isomorphic Labs plans to use AI software to create new drugs and medicines. Demis Hassabis, the CEO and co-founder of DeepMind, is also the founder and CEO of Isomorphic Labs and will remain CEO of D both the companies. Demis is also a UK Government AI Advisor since 2018.

Headquartered in London, Isomorphic Labs is a separate entity from DeepMind and has its own dedicated resources.

Interestingly, Isomorphic Labs will eventually pick up a new CEO as the company moves into later stages. The newly formed company is looking to hire computational and medicinal chemists as well as machine learning experts in addition to filling scientific, engineering and operational roles.

Last year DeepMind’s breakthrough AI system AlphaFold2 was recognized as a solution to the 50-year-old grand challenge of protein folding, capable of predicting the 3D structure of a protein directly from its amino acid sequence to atomic-level accuracy. AlphaFold is touted as the most important datasets since the mapping of the Human Genome.

Google Creates AI That Can Do Things Which Previously Only Humans Could

Tech giant Google's artificial intelligence (AI) division, DeepMind, which the company acquired back in 2014, is currently working on a project that can change the course of AI for years to come.

The division, which claims to be on a scientific mission to push the boundaries of AI, developing programs that can learn to solve any complex problem without needing to be taught how, recently shared that they have mastered an AI that can make its own plans.

According to DeepMind, they have successfully created “Imagination-Augmented Agents” that are capable of “imagining” the possible consequences of their actions, and interpret those simulations accordingly. The London-based company claims that these agents can make the right decision for what it is they have set out to achieve.

DeepMind researchers shared that in a number of tasks that were carried out to test these agents, they ended up handsomely outperforming baseline agents. According to the researchers, these agents are the closest that AI has come to human thinking. They think like humans, try out different strategies in their head about a situation prior to executing it, and are therefore able to learn despite having little or no real life experience.

Talking about their path-breaking invention, DeepMind researchers in a blogpost divulged that their Imagination-Augmented Agents comes with an 'imagination encoder’- a neural network which learns to extract any information which might be useful for the agent’s future decisions, but ignore which is not relevant. They learn to interpret their internal simulations, which allows them to use models which coarsely capture the environmental dynamics, even when those dynamics are not perfect. They also make use of their imagination efficiently by adapting the number of imagined trajectories to suit a particular problem. On the top of it all, these agents can learn different strategies to construct plans by withers choosing between continuing a current imagined trajectory or restarting from the very beginning.

In order to assess the performance of its agents, DeepMind tested them on a spaceship navigation game and the famous puzzle game Sokoban, both of which require forward planning and reasoning to win.

According to the researchers, the agents tested out well for both the tasks. In fact, they outperformed the imagination-less baselines considerably, meaning they learned with less experience and are able to successfully cope with the imperfections in modelling the environment. This is because they're able to extract more knowledge from internal simulations.

In the near future, DeepMind wants to create computers that can survive in complex environments where unpredictable problems can arise anytime.