Do you think there is a real future for AI in decarbonization technologies? Is anyone using this today to reach net-zero goals?

The simple, but not helpful answer to your question, is yes. There are very few domains involving optimization of resources that are not already benefiting from the application of a range of AI techniques: advanced machine learning through to cognitive compute. Eg. Google DeepMind has been applied to the optimization of power consumption in data centers and IBM Watson to the optimization of wind turbines. However, as with all energy consumption strategies it will be important to evaluate the total economic cycle to determine whether the reduction in consumption outweighs the net increase in consumption required to power these AI techniques.

That is a great question.  I was recently at a Deloitte presentation where this came up.  Data centers are energy consumers, and while we do not have enough green power to power them today, that means we will have to use natural gas until alternative (small form factor nuclear) energy sources are available.  But, according to Deloitte, data center energy usage will grow to account for around 3%  (from sub 1% today) of total green house gas emissions. Even with the growth in AI.  And once they use nuclear power, they will emit very few green house gases.  Also, next gen AI chips will again be more efficient power wise.  This means if you can identify where AI saves other areas of energy, then you don't have to worry about the data center processing emissions, much, in your calculations.  Not to mention the improvements in calculations and optimizations for all things, including trucking routes, energy grids, and even green house gas estimates.