Neuroscientists and military vets: the inner workings of the team that ‘hacks’ Microsoft’s AI tools before their public debut

english.elpais.com·Patricia Fernández de Lis
View original article
0out of 100
Moderate — some persuasion patterns present

This article aims to reassure readers about Microsoft's ethical AI development by highlighting its internal 'red team' and security protocols. It effectively uses quotes from high-ranking officials like President Brad Smith and engineer Ram Shankar Siva Kumar to present Microsoft as a responsible innovator that proactively addresses potential harms and even supports companies like Anthropic against military contracts. While it showcases Microsoft's internal efforts, it minimizes discussions about external regulatory oversight or the broader ethical challenges of AI in warfare.

FATE Analysis

Four dimensions of psychological manipulation: how content captures Focus, exploits Authority, triggers Tribal identity, and engineers Emotion.

Focus3/10Authority4/10Tribe2/10Emotion3/10
FFocus
0/10
AAuthority
0/10
TTribe
0/10
EEmotion
0/10

Focus signals

novelty spike
"But applying it to generative artificial intelligence is something relatively new, and Microsoft is attributed with being a pioneer in the area, having formed its team in 2018."

This highlights the 'newness' and 'pioneering' aspect of Microsoft's work, suggesting uncharted territory and inherent interest.

attention capture
"Microsoft president Brad Smith takes a moment to reflect before using the word “guardrails” with the ease of someone who has given a great deal of thought to the dangers of the abyss."

The dramatic imagery of 'dangers of the abyss' and the president's 'reflection' acts as a hook to draw the reader into the narrative and emphasize the gravity of the topic being discussed by Microsoft.

Authority signals

credential leveraging
"Microsoft president Brad Smith takes a moment to reflect before using the word “guardrails” with the ease of someone who has given a great deal of thought to the dangers of the abyss."

The title 'Microsoft president' lends significant weight to Smith's words, implying deep knowledge and serious consideration of the topic.

institutional authority
"Microsoft, in fact, has supported Anthropic in its fight."

Microsoft's support for Anthropic against the Pentagon leverages the image of Microsoft as a powerful, principled corporation standing against military power, implicitly endorsing Anthropic's stance.

expert appeal
"“Before a product is launched, the red teams break the technology so that others can rebuild it to be more solid and secure,” explains Ram Shankar Siva Kumar, who self-identifies as a “data cowboy” and is the leader of the red team."

Kumar's role as 'leader of the red team' and his description as a 'data cowboy' establishes him as an expert in the field, lending credibility to his explanations.

credential leveraging
"Along with Kumar, the red team’s operations are co-directed by Tori Westerhoff, whose background combines cognitive neuroscience — she studied at Yale and was one of the first members of the Wharton Neuroscience Initiative — and national security strategy, having worked at intelligence and defense agencies."

Westerhoff's prestigious educational background (Yale, Wharton) and experience in intelligence and defense agencies are invoked to establish her significant expertise and authority in the field, thereby bolstering the credibility of the red team's work.

expert appeal
"This way of seeing AI aligns with the vision of Mustafa Suleyman, one of the founders of Deepmind (now part of Google) and CEO of Microsoft."

Citing Suleyman's perspective, given his high-profile roles and expertise in AI, adds significant weight and validation to the article's points about responsible AI development.

Tribe signals

us vs them
"Just a few days ago, it was made public that artificial intelligence firm Anthropic has sued the Pentagon for blacklisting it after the company turned down a contract for the defense entity to utilize its technology. It is the current debate that is raging in the world of Big Tech, and a very familiar issue at Microsoft."

This sets up a subtle 'us vs. them' dynamic between ethical tech companies (Anthropic, implicitly Microsoft) and the military/defense complex (Pentagon), aligning the reader with the former.

Emotion signals

fear engineering
"Microsoft president Brad Smith takes a moment to reflect before using the word “guardrails” with the ease of someone who has given a great deal of thought to the dangers of the abyss."

The phrase 'dangers of the abyss' evokes a sense of potential catastrophe or severe negative consequences, subtly engineering a sense of fear or apprehension about uncontrolled AI.

urgency
"His AI internal affairs team has already analyzed more than 100 of the company’s products. Microsoft does not release information regarding how many people work in the team, nor how many or which products whose release they have halted. But he does say that the team has the power to do so: “No high-risk AI system is implemented before undergoing an independent test. If our team identifies serious risks that have not been mitigated, the product is not released until those problems are resolved,” says Kumar."

The implied magnitude of halted products due to 'serious risks' attempts to create a sense of urgency about AI safety and the crucial role of the red team in preventing potential harm. The phrase 'not released until those problems are resolved' emphasizes the gravity of the issues.

fear engineering
"This way of seeing AI aligns with the vision of Mustafa Suleyman, one of the founders of Deepmind (now part of Google) and CEO of Microsoft. A few days ago, he wrote in the publication Nature that an apparently conscious AI could become a weapon."

The idea of 'an apparently conscious AI could become a weapon' directly taps into anxieties about out-of-control AI and existential threats, engineering fear.

Narrative Analysis (PCP)

How the article reshapes thinking: Perception (what beliefs are targeted), Context (what information is shifted or omitted), and Permission (what behavior is being encouraged).

What it wants you to believe

The article aims to install the belief that Microsoft is a highly responsible and ethical developer of AI, proactively addressing potential harms and dangers associated with its technology through rigorous internal processes and a diverse, expert team. It wants the reader to believe that Microsoft's 'guardrails' are robust and effective.

Context being shifted

The article shifts the context of AI development from one that might require significant external ethical oversight or government regulation to one where internal corporate responsibility (Microsoft's 'red team' and principles) is presented as the primary and sufficient mechanism for ensuring ethical AI. This makes Microsoft's self-governance appear adequate.

What it omits

The article omits broader context regarding the independence and actual oversight power of Microsoft's 'red team' in relation to profit motives or executive pressure to release products. It also significantly downplays or omits the role of external regulatory bodies, independent ethical review boards, or legislative efforts in controlling AI development, focusing solely on internal corporate mechanisms. The article also mentions the Anthropic lawsuit against the Pentagon and Microsoft supporting Anthropic, but frames this as part of a 'debate' without elaborating on the military-industrial complex's significant influence on tech companies or the ethical dilemmas many tech workers face regarding defense contracts; instead, it uses it to highlight Microsoft's seemingly ethical stance.

Desired behavior

The reader is nudged to feel reassured and confident in Microsoft's approach to AI development, granting implicit permission for Microsoft (and by extension, other large tech companies adopting similar rhetoric) to continue innovating rapidly with AI largely under their own internal ethical frameworks, without necessitating strong external governmental or public oversight.

SMRP Pattern

Four manipulation maintenance tactics: Socializing the idea as normal, Minimizing concerns, Rationalizing with logic, and Projecting blame.

-
Socializing
-
Minimizing
-
Rationalizing
-
Projecting

Red Flags

High-severity indicators: silencing dissent, coordinated messaging, or weaponizing identity to shut down debate.

-
Silencing indicator
!
Controlled release (spokesperson test)

"Smith answers, “We have principles, we define them and we publish them. By definition, those principles create guardrails. And we stay in our lane within them. It’s not just about when we should use technology, but also about when we shouldn’t use it.”; “No high-risk AI system is implemented before undergoing an independent test. If our team identifies serious risks that have not been mitigated, the product is not released until those problems are resolved,” says Kumar."

-
Identity weaponization

Techniques Found(6)

Specific propaganda techniques identified using the SemEval-2023 academic taxonomy of 23 techniques across 6 categories.

Loaded LanguageManipulative Wording
"Microsoft president Brad Smith takes a moment to reflect before using the word “guardrails” with the ease of someone who has given a great deal of thought to the dangers of the abyss."

The phrase 'dangers of the abyss' disproportionately heightens the perceived threat of AI, framing it as an existential peril rather than a complex technological challenge, thus using emotionally charged language to implicitly cast Microsoft as a protector.

Name Calling/LabelingAttack on Reputation
"His AI internal affairs team"

Labeling the team as 'AI internal affairs' evokes connotations of internal investigations and potential wrongdoing, subtly questioning the team's role in a non-neutral way regarding AI development.

Loaded LanguageManipulative Wording
"“AI can generate problems from security failures to psychosocial damage. People use Copilot [Microsoft’s AI] in moments of great vulnerability, so observing how these systems can fail before they get to the user is fundamental,” he says."

The phrase 'moments of great vulnerability' is emotionally charged, designed to evoke empathy and concern, thereby magnifying the perceived necessity and importance of Microsoft's 'red team' work in preventing harm.

Exaggeration/MinimisationManipulative Wording
"Westerhoff believes, in fact, that only the human mind is capable of “imagining the spaces that have not yet been observed, that are not completely defined or explored. Our work consists of innovating and creating beyond the space that has been systematized.”"

This statement exaggerates the uniqueness of human cognitive ability in the context of AI development, portraying the human team's role as almost mystical and irreplaceable, potentially downplaying the capabilities of AI in certain exploratory tasks.

Loaded LanguageManipulative Wording
"A few days ago, he wrote in the publication Nature that an apparently conscious AI could become a weapon."

The phrase 'apparently conscious AI' and the immediate pivot to it becoming a 'weapon' uses emotionally charged and speculative language to create a sense of fear and urgency, positioning Microsoft and its experts as crucial in preventing this dire outcome.

SlogansCall
"“responsible AI is not a filter applied at the end of development, but a foundational part of the process,”"

This is a catchy, concise phrase that encapsulates Microsoft's stated philosophy on AI development, serving as a memorable slogan to promote their approach.

Share this analysis