Thread by @RikeFranke

France has published its military #AI strategy! Very exciting. Will read today.

defense.gouv.fr/salle-de-press…

Also, @BMVg_Bundeswehr, how's it looking? Any chance we might see a similar strategy from you?

@BMVg_Bundeswehr Alright people, strap in, let's do this.
Thread on France's new military #AI strategy.
(note - I couldn't find an English translation, so I'll translate, but keep in mind the translation isn't official).
[1]

The strategy starts pretty broad, referring to the "beginning of AI" in 1956 (i.e. Dartmouth conference) and notes that AI these days is simultaneously seen as the hope for humanity or the source of its destruction. [2]

Very early on, the strategy notes that there is an "intensive, global competition" going on. And then, in bold: "In this race, the stakes and the pace are such that any stall would be irremediable."
So basically: France says, 'there is an #AI race - let's not lose it. [3]

We then dive into #AI more specifically, which the strategy calls "a revolution". It also notes that for the military, AI isn't a goal in itself, but it's about helping the French armed forces to do their job better. [4]

In this section, the strategy makes an important point. Namely, it notes that there are important #AI advances in the civilian sector, driven primarily by US and Chinese firms. This can benefit the military realm, because, also the military can use better, say, AI-enabled HR. [5]

BUT: When it comes to *actual military* systems, a lot of the civilian AI developments are only so useful. They may not be able to deal with military data (infrared images, radar data), may not be sufficiently robust, and need to work in areas that are unknown, etc.[6]

Hence: the military can benefit to some extent from commercial #AI, but the more specifically military it becomes, the more there is a need for own developments, and "autonomy".
[7]

This was the introductory part, now to the real meat, i.e. what #AI may, according to the French military thinkers, may actually do in a military conflict.

There is little ambiguity, the strategy says French military decision-makers need to take advantage of #AI to "gain in velocity, get better reconnaissance and target detection ... do faster&better targeted actions". All this "while guaranteeing the respect for the laws of war".

Getting more specific, the strategy first and foremost emphasises that AI can increase the speed of operations by basically speeding up and optimising everything: data and threat analysis, decision-making and more. [9]

#AI can help protect the soldiers, the strategy notes - by improving/optimising health care, improving training (VR, simulation) and robotisation. [10]

This bit a lot of people won't like:
the strategy states that the combatant may also be better protected because "AI will bring significant support for the respect of International Humanitarian Law". How so?
[11]

"#AI will strongly contribute to improve the discrimination between combatants and non-combatants",
"improve proportionality",
and "guarantee that an action [is] determined by strict need." [12]

"So, contrary to some conventional wisdom, #AI, well supervised and controlled, has the potential to lead to better consideration by the French armed forces of the fundamental principles of the law of armed conflict." [13]

Side note: this argument (AI will help with IHL) isn't new. But in my perception, this is shared by less experts today than it was a few years ago. Also, what strikes me in this document is that it's formulated in pretty certain tones, "AI will do this", rather than "AI may"

Let's continue. The strategy also notes that #AI can help with repetitive tasks and maintenance of material. But now we turn to risks, with the strategy noting that as #AI may be everywhere, risks may as well. [14]

Risks as noted here are the predictability of French action by #AI-enabled enemy forces, paralysation of the system, #AI-enabled disinformation and fake news, #cyber attacks.

The strategy also warns that there is an #AI arms race happening now and many states believe they can use AI to modify the global military hierarchy to their advantage.

Finally, the strategy notes that #AI so far is not yet very robust and can be manipulated. It also notes some problems inherent to machine learning, such as bias, ad the need for good data.

You're all still there? We're now into the section on the international competition on #AI. In the writers' view, the landscape is as follows: two superpowers (US and China) "outside anyone's reach" that have the data, the ecosystem, the companies. [18]

Then, "an intermediate power in the making", i.e. the European Union. Very interesting assessment here: "The EU's strict positioning on legal-ethical issues may be a strength or a like a weakness, depending on its impact". Strength if it becomes a normative power that...
[19]

...gets many other players on board, weakness if an excessive regulation leads to timorous* commercial development or hinders research. [20]

*not gonna lie, I had to look up this word in *both* languages 😅

And then there is a second circle of states that includes France as well as Germany, the UK, Japan, South Korea, Singapore, Israel and Canada.
They have some capabilities but will need to cooperate and maximise their strengths in some niche capabilities. [21]

Side note: I am a big fan of this "let's see the world as it is" geopolitical approach that France adopts. You may not agree with everything in this strategy, but so far there is rather little political spinning and more of a realist approach.

Noteworthy is the emphasis on "preserving sovereignty". The strategy notes that the whole #AI sector depends too much on US firms, and that the lead by the US and China is strong. It states that France cannot be dependent on technologies that it does not master. [22]

France, the strategy says, is good on fundamental research, but less on real-world adoption of #AI. That needs to change, and data collection needs to be improved (mirrors the national AI strategy published a while back). [23]

We now have 20 pages of recommendations & plans for the minister - no point in listing them all. I will go through and then highlight those parts I find noteworthy.
In the meantime, read for yourself here: defense.gouv.fr/english/salle-…
#France #AI #AIStrategy #Europeansovereignty

Thread continues here

https://twitter.com/RikeFranke/status/1174594529476956160