Guest essay by Eric Worrall
When the awakening awoke, the awakening awoke. Back in December, Google dismissed Timnit Gebru, co-head of the AI Ethics Unit, in relation to her work “On the dangers of stochastic parrots: Can language models be too big?”. Google just fired their other ethics chief Margaret Mitchell for apparently trying to gather evidence while investigating Timnit’s fall.
I’m fired: Google AI in the meltdown when the co-lead of the ethics department was pushed out of office just a few weeks after the colleague was overthrown
Plus: IBM is reportedly trying to sell Watson AI Health and more
Katyanna Quach Mon 22 Feb 2021 // 12:21 UTC
Google has completed its investigation into the controversial fall of Timnit Gebru, co-head of its ethical AI division. The advertising giant promised to introduce new procedures for “potentially sensitive employee exits” but did not publish its results.
Gebru said she was fired for warning employees in an internal memo that promoting diversity, equality and inclusion within the Silicon Valley Goliath was a waste of energy due to management apathy. Google claimed she effectively resigned.
…
Meanwhile, Margaret Mitchell, who ran the Ethical AI unit alongside Gebru, said on Friday that she had been fired. Mitchell had been locked out of her corporate account for weeks.
…
Read more: https://www.theregister.com/2021/02/22/in_brief_ai/
From the newspaper “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?”;
…
3 ENVIRONMENTAL AND FINANCIAL COSTS
Strubell et al. Recently, the US dollar model training and development costs and estimated 𝐶𝑂2 emissions were compared [129]. While the average person is responsible for an estimated 5 t 𝐶𝑂 2 𝑒 per year, 2 the authors trained a transformer model (large) [136] with neural architecture seek and estimate that the training method emitted 284t of 𝐶𝑂2. It has been estimated that training a single BERT base model (without hyperparameter tuning) on GPUs requires as much energy as a trans-American flight.
While some of this energy comes from renewable sources or cloud computing companies use carbon credit offset sources, the authors note that most of the energy from cloud computing providers does not come from renewable sources and many energy sources around the world are not carbon neutral. Furthermore, renewable energy sources are still costly to the environment 3 and data centers with increasing computational effort are taking away other potential uses of green energy 4, underscoring the need for energy efficient model architectures and training paradigms.
…
Read more: http://faculty.washington.edu/ebender/papers/Stochastic_Parrots.pdf
Timnit has also criticized other issues with AIs. For example, in 2018, she helped halt the rollout of an Amazon facial recognition system used by law enforcement agencies by demonstrating that the flawed Amazon system was 34% less able to correctly identify black women than its ability to correctly identify white men to identify. The problem – the dataset used to train the AI contained mostly white faces.
What can I say – losing an ethics head could be an accident. Losing two in quick succession looks like carelessness and may even raise suspicions that Google really wants a compliant ethics team that does whatever top management demands of it.
Like this:
Loading…
Comments are closed.