Yesterday, AIM published an article on how difficult it is for the small labs and individual researchers to persevere in the high compute, high-cost industry of deep learning. Today, the policymakers of the US have introduced a new bill that will ensure deep learning is affordable for all.
The National AI Research Resource Task Force Act was introduced in the House by Representatives Anna G. Eshoo (D-CA) and her colleagues. This bill was met with unanimous support from the top universities and companies, which are engaged in artificial intelligence (AI) research. Some of the well-known supporters include Stanford University, Princeton University, UCLA, Carnegie Mellon University, Johns Hopkins University, OpenAI, Mozilla, Google, Amazon Web Services, Microsoft, IBM and NVIDIA amongst others.
Sign up for your weekly dose of what's up in emerging technology.
The objective of this Act is to establish a task force that develops a roadmap for a national AI research cloud.
About The Bill
The volatile relations with China and alleged espionage of intellectual property made the US uncompromising in its pursuit to retain the top spot as the hub of innovation. The US has even set up a National Security Commission on Artificial Intelligence (NSCAI), and the current Task Force Act is an endorsement of the same.
The NSCAI, co-chaired by Eric Schmidt, the former chief executive of Alphabet (Google’s parent company), and Robert Work, the former deputy secretary of defence, is a congressionally mandated, independent federal commission set up last year. The objective of this new task force act is to establish a national resource that would accelerate and strengthen AI research across the U.S. by removing the high-cost barrier to entry of compute and data resources.
The success of which would democratise AI research outside of elite universities and large technology companies and will give academic researchers the tools needed to advance AI far into the future. This, in turn, would give colleges and universities an edge to remain competitive in AI.
The creation of the U.S. Government-led task force for AI was initiated by Stanford HAI earlier this year in March. The Stanford researchers called for a nationwide collaboration from academia, government, and industry to establish a National Research Cloud.
Now it’s up to the Congress to contrive a plan with the help of technical experts across academia, government, and industry, as to how the U.S. can build, deploy, govern, and sustain a national AI research cloud.
Why Does The Bill Matter
“As AI and ML become increasingly core to all of our lives, the National AI Research Resource Task Force Act of 2020 is an important part of ensuring that the internet remains open and accessible to all.”-Jofish Kaye, Principal Scientist at Mozilla
Smaller organisations do not have the resources to replicate the successes of larger organisations. We witnessed last month how Uber was forced to wind down its AI labs due to unfavourable circumstances. The R&D aspect of AI doesn’t look commercially viable in the short run, hence the smaller research groups will have no other way than to rely on the APIs released by the likes of OpenAI.
Since 2012, AI has made remarkable progress on a broad range of capabilities including object recognition, game playing, machine translation, and more. The training cost increased over time for state-of-the-art deep learning models starting with AlexNet in 2012 to AlphaZero in 2017. According to a study conducted by Allen Institute, there has been an overall increase of 300,000x, with training cost doubling every few months. An even sharper trend can be observed in case of NLP models such as BERT and more recently with OpenAI’s GPT-3.
Companies like Google and Microsoft have invested massively in developing new models. They even own the data centres that are used for cloud computing. But to those who have to buy to build, it’s an uphill task.
The cloud computing cost of some of the widely used models, as can be seen above, depicts the harrowing tale of individual researchers who cannot afford to toy around with their models. Innovations are indifferent to financial hassles. Now with the Task Force Act, the US government is determined to unburden the researchers of their initial hurdles.