How AI and advanced computing can bring us back from the brink of accelerating climate change

Barely a week goes by without another dramatic report that humanity and the planet are reaching a tipping point in climate change. The latest report was a startling analysis from the World Meteorological Organization and a startling criticism from the UN Secretary-General. Both were shared on the last day of April.

Artificial intelligence will determine whether we cross the tipping point or pull back from the brink.

AI is one of the key tools left in the fight against climate change. AI is focused on risk prediction and prevention of harmful weather events such as wildfires and carbon offsets. This is said to be essential to ensure that companies meet his ESG goals.

But it is also a facilitator. AI requires enormous computing power, and energy is expended when designing algorithms and training models. And just as software has consumed the world, so will AI.

AI is expected to contribute $15.7 trillion to the global economy by 2030, more than the GDPs of Japan, Germany, India and the UK. From composing emails and writing code using ChatGPT to creating art using a text-to-image conversion platform, many are using AI as ubiquitously as the Internet. there are people

The power AI uses has grown over the years. For example, the power required to train the largest AI model doubles roughly every 3.4 months, and he increased by a factor of 300,000 between 2012 and 2018.

This expansion provides an opportunity to solve major real-world problems in everything from security and healthcare to hunger and agriculture. It will also have punitive effects on climate change.

high energy cost

Computing goes hand in hand with higher energy costs and a larger carbon footprint, which contributes to global climate change.

This is especially true for AI. Huge numbers of GPUs running machine learning algorithms get hot and need to be cooled. Otherwise it will melt. Training a single large language model (LLM) has a high carbon footprint and requires a dizzying amount of energy.

for example:

As we move into the GPT4 era and our models get bigger, so does the energy required to train them. GPT-3 was 100 times more than his GPT in the previous generation, GPT-4 was 10 times more than his GPT-3. Meanwhile, larger models are being released sooner. GPT-4 arrived in March 2023, almost four months after the release of his ChatGPT (powered by GPT-3.5) at the end of November 2022.

To balance things out, we shouldn’t assume that the carbon footprint of AI will continue to grow as new models and companies emerge in space. Her Geeta Chauhan, an AI engineer at Meta, uses open source software to reduce LLM’s operational carbon footprint. Her latest research shows her 24x reduction in carbon emissions compared to GPT-3.

But the popularity and exponential power of AI is undermining much of the current climate action, calling into question its potential to be part of the solution.

We need solutions that enable AI to thrive while keeping our carbon footprint in check. What should I do?

Relieve carbon poisoning

As always, technology will help us out of this predicament.

For the AI ​​explosion to be sustainable, advanced computing must come to the fore and do the heavy lifting of many tasks currently performed by AI. The good news is that we already have advanced computing technologies that can perform these tasks more efficiently and faster than AI. Another advantage is that the energy used can be significantly reduced.

In short, advanced computing is the most effective tool we have for mitigating AI’s carbon poisoning. By doing so, we can slow the progress of climate change.

A number of different advanced computing technologies are emerging that can solve some of the problems AI currently tackles.

For example, quantum computing outperforms AI in drug discovery. As human lifespans have increased, we have encountered more and more complex and untreatable new diseases. It’s called the “Better Than The Beatles” problem, and the new drug is a slight improvement on an already successful treatment.

Historically, drug development has focused on rare events in datasets and made educated guesses to design appropriate drugs that target and bind disease-causing proteins. Efficient use of LLM can assist in this task.

LLM is very good at predicting which words in the vocabulary best fit a sentence to accurately convey meaning. Drug discovery is not very similar, as it is a matter of identifying the optimal match or configuration of molecules within a compound to achieve therapeutic results.

However, since molecules are quantum elements, quantum computing is much better at tackling this problem. Quantum computing has the ability to rapidly simulate vast numbers of binding sites within pharmaceuticals to create suitable configurations for treating currently incurable diseases.

Advanced Computing: Quantum and Beyond

Quantum’s capabilities mean these can be solved much faster and with less energy usage.

Another development that could really enhance AI is photonics, or so-called optical computing, which uses laser-generated light instead of electricity to transmit information.

Some companies are building computers using this technology. This technology is more energy efficient than most other computing technologies and is increasingly being recognized as a means of achieving net zero.

Elsewhere we have neuromorphic computers. It is a form of computer engineering in which elements of a computer system model those of the human brain and nervous system. They perform computations to recreate the analogue nature of our nervous system. Experiments with this technology include projects by Mythic and Semron. Neuromorphic is another eco-friendly option that requires more investment. Its hardware has the potential to run large-scale deep learning networks that are more energy efficient than comparable traditional computing systems.

For example, information processing by 100 billion neurons consumes only 20 watts, which is about the same as an energy efficient light bulb in your home.

Developing and applying these innovations is essential to curbing climate change.

Advanced Computing Leader

There are plenty of startups (and investors) around the world who are crazy about advanced computing, but only a handful are focused on so-called impact areas like healthcare, the environment and climate change. .

The most exciting companies developing energy and drug discovery use cases in the field of quantum computing are Pasqal (whose co-founder won the Nobel Prize in Physics in 2022), Qubit Pharmaceutical, and IBM. When it comes to photonics, it sees Lightmatter and Luminous as the world’s most influential leaders, while in neuromorphic computing it tracks progress from Groq, Semron, and Intel.

Advanced computing is essential to achieving the energy efficiency needed to combat climate change. Running artificial neural networks on GPUs takes too long and consumes too much energy.

By embracing advanced computing methods as an alternative to AI, companies can significantly reduce the impact of AI on the environment, while using their enormous capabilities to predict some of the impacts of climate change, such as predicting wildfires and extreme weather events. We can assure you that we can mitigate the

We are approaching the end of existence for our environment. But the situation is not hopeless.

Adopting advanced computing is one of the most reliable and powerful resources for combating this problem. We need to invest in these technologies now to solve the greatest challenges facing humanity.

Francesco Ritchiti VC at Runa Capital.