Thread by @HeikoBorchert

Recommended #weekendread: #AI and #National #Security #NatSec US Congressional Research Service via @saftergood
fas.org/sgp/crs/natsec…
cc @Dr_JMRickli @Ce_Moll @WachterBDI @RikeFranke @MoehringJohanna @M__Verbruggen

The following two charts are particularly interesting:
1. #Defense #DoD spending on #AI: 2012-2017 (page 5)

2. #Chinese #investment in US #AI companies: 2010-2017 (page 20)

If you read but one chapter of this informative report, go straight to "#AI #acquisitions #challenges." The chapter discusses 8 major #hurdles/#obstacles that need to be tackled if #DoD – or any other #MoD!!! – wants to successfully use #AI

I. The #combat #environment is #unstructured and #nonbenign thus requiring #adapations of #AI #algorithms. BUT: Commercial companies may not have strong #incentives to create AI of this type "because most of their tools will encounter much less #contested situations."

II. #Standards of #safety & #performance may be different for #defense applications as #tolerances in a combat environment are likely to "deviate from failure rates deemed acceptable for a #civilian AI application." – Think abt the impact of different #cutural #predispositions!

III. "No independent entity in the commercial sector or inside government is charged with #validating #AI system performance and #enforcing #safety standards." – Consider the #certification challenge with current defense systems and compare this to #AI!

IV. Existing #procurement procedures are "not agile enough for past-paced software systems like AI". Consider this:
a. DoD Instruction 5000.02 on acquisition requires about 91 months from initial analyses to #IOC
b. Commercial companies typically develop AI software in 6-9 months

V. Disconnect between #defense and #commercial #companies on the #incentives for #collaboration in particular with regard to corporate interests in near-term #profits & growth and "insistence on preserving #IP"

VI. "Companies are refusing to work with DoD altogether because of #concerns over #AI being used for #government #surveillance and #lethal #applications." See Alphabet's withdrawal of BostonDyamics from DARPA contests

As with every kind of #innovation, the remaining two #hurdles are classical examples of #organizational #resistance and #reluctance to change:

VII. "The #culture within the defense sector itself may create an #impediment to #AI #integration. (...) Members of project #MAVEN have reported a resistance to change b/c the #disruption that comes with #AI integration does not provide intuitive benefit." LET THIS SINK IN!

VIII. There might be a #lack of #conceptual #creativity as DoD could "simply use #AI to improve existing processes instead of #capitalizing on the tech's potential to produce a more significant benefit on the battlefield."

CONCLUSION: All of this suggests that #armed #forces will need to go a long way to fully #embrace, #integrate and #adapt to #AI along the #DOTLMPFI lines of #capability development – talk about the #AI #arms #race is premature

FYI @mchorowitz @oth_mdos @wavellroom @Strategy_Bridge @JulesHawke @Strikepod @WarintheFuture @WarProf @LandSharkUK @flwaustralia @_schuetzt @qmacbill @davidblogistics @MLC_Haas @Mauro_Gilli @aa_gilli @Olivier1Schmitt @EBKania @paul_scharre @bvoelkl @autonomy_tommy