Atlas of AI, book review: Mapping out the total cost of artificial intelligence


Atlas of AI: Electricity, Politics, and the Planetary Fees of Synthetic Intelligence • By Kate Crawford • Yale University Push • 336 web pages • ISBN: 978–three hundred-20957- • £20   

“Request forgiveness, not authorization” has lengthy been a guiding theory in Silicon Valley. There is no technological area in which this theory has been a lot more practiced than the device learning in modern AI, which is dependent for its existence on big databases, almost all of which are scraped, copied, borrowed, begged, or stolen from the big piles of facts we all emit day by day, knowingly or not. But this facts is barely at any time rigorously sourced with the subjects’ authorization.  

“Mainly because we can,” two sociologists convey to Kate Crawford in Atlas of AI: Electricity, Politics, and the Planetary Fees of Synthetic Intelligence, by way of acknowledging that their tutorial institutions are no distinct from technological know-how companies or authorities companies in pertaining to any facts they locate as theirs for the taking to teach and test algorithms. Pictures develop into infrastructure. This is how device learning is designed. 

Anyone wishes to speak about what AI is good or dangerous for — pinpointing facial illustrations or photos, interpreting speech commands, driving cars (not nevertheless!). Many want to pour ethics around present-day AI, as if generating guidelines could change the army funding that has described its essential mother nature. Couple of want to focus on AI’s legitimate expenditures. Kate Crawford, a senior researcher at Microsoft and a research professor at the University of Southern California, is the exception. 

SEE: Creating the bionic brain (totally free PDF) (TechRepublic)

In Atlas of AI, Crawford starts by deconstructing the famed rivalry that ‘data is the new oil’. Typically, that leads folks to speak about data’s financial price, but Crawford focuses on the point that equally are extractive technologies. Extraction is mining (as in ‘data mining’ or oil wells), and where by mining goes, so adhere to environmental damage, human exploitation, and profound modern society-large penalties.  

Crawford underlines this level by heading to Silver Peak, Nevada, to go to the only operating lithium mine in the US. Lithium is, of study course, a important part in battery packs for everything from smartphones to Teslas. Crawford follows this up by thinking about the widening implications of extraction for labour, the sources of facts, classification algorithms, and the nation-condition behaviour it all underpins, ending up with the ability constructions enabled by AI-as-we-know-it. This way lies Project Maven and ‘signature strikes’ in which, as previous CIA and NSA director Michael Hayden admitted, metadata kills folks. 

Snake oil

Nonetheless some of this is patently phony. Crawford traces again the image datasets on which the hottest disturbing snake oil — emotion recognition — is based mostly, and finds they ended up created from posed pics in which the subjects ended up instructed to deliver exaggerated examples of psychological reactions. In this case, ‘AI’ is manufactured all the way down. Is there, as Tarleton Gillespie questioned about Twitter tendencies, any true human reflection there? 

While other technological know-how books have tackled some of Crawford’s subject areas (far too many of which have been reviewed listed here to listing), the closest to her built-in structural technique is The Fees of Connection by Nicholas Couldry and Ulises A. Mejias, which views our current technological reconfiguration as the beginnings of a new romantic relationship in between colonialism and capitalism. 

“Any sufficiently advanced technological know-how is indistinguishable from magic,” Arthur C. Clarke famously wrote. Next Crawford, this seems to be a lot more like: “Any technological know-how that seems to be like magic is hiding one thing.” So many dim techniques lie in how the sausage is designed. 

Recent AND Similar Written content

The Fees of Connection, guide evaluate: A wider watch of surveillance capitalism

The most important expense in database record, the most important social community at any time, and other graph tales from Neo4j

Now Google is employing AI to structure chips, considerably more rapidly than human engineers can do the occupation

Drones with exoskeletons experience off in soccer challenge

Frightened, human? Emotion-detection AI meets eye-tracking technological know-how

Read a lot more guide opinions