AI Influence Ops Exposed, Siri’s AI Update

31 May 2024: OpenAI battles AI misuse, Google’s search errors, Apple’s Siri overhaul, Perplexity AI’s new feature, US restricts AI chips, ACLU targets bias.

May 31, 2024

three men and one woman laughing during daytime

Good Morning & Happy Friday,

Latest in AI,

1. 1-Minute Roundup
2. Headlines
3. AI Investment Deals
4. Infographic Of The Day
5. Others

1-Minute Roundup

Headlines:

OpenAI halts five covert influence operations misusing AI tools for misinformation.
Why Google's AI search feature faces criticism for inaccuracies and bizarre recommendations.

AI Investment Deals:

ifeel raises $20M for AI mental health solutions.
Maven AGI secures $20M for AI customer support.
EthonAI obtains $16.5M for AI manufacturing improvements.
Bootloader Studio gets $5M for affective AI and spatial computing.
Maxa raises $21M for Snowflake Native Apps and ERP AI insights.
Zendata receives $2M for AI governance and data privacy.

Others:

The new feature generates detailed reports from single prompts using real-time web searches.
License delays for AI chip exports to the Middle East due to security concerns.
Complaint filed against Aon for AI hiring tools allegedly biased against disabled and minority candidates.
Apple will enhance Siri with AI capabilities to control individual app functions.

Share Artificial Intelligence and Machine Learning

Headlines

OpenAI Shuts Down Influence Networks Misusing Its AI Tools

an abstract image of a sphere with dots and lines

OpenAI reports identifying and shutting down five covert influence operations across multiple countries that were improperly accessing its AI models, including ChatGPT, to manipulate public opinion while obscuring their identities.

Why it Matters: The report underscores emerging threats around AI being weaponised by bad actors to mislead the public and interfere in democratic processes like elections at scale.

Zoom In:

Groups in Russia, China, Iran, and Israel were flagged for leveraging OpenAI tools alongside traditional misinformation tactics.
Russian "Bad Grammar" operation used AI for coding bots, generating comments supporting anti-Ukraine narratives.

Backstory: OpenAI defined the targets as covert attempts to manipulate public opinion by obscuring their true identities and intentions.

Big Picture: As generative AI capabilities advance, technology companies must proactively detect and mitigate emerging threats around coordinated campaigns spreading AI-generated disinformation.

Reality Check: OpenAI assessed the identified campaigns as having limited reach but warned against complacency as influence operations evolve tactics over time.

How it Works: Influence groups combined AI text/image generation with traditional misinformation methods like memes and manually written posts to amplify their messages.

What's Next: OpenAI aims to share indicators with peers to enable broader detection, promising more transparency reports on emerging AI threat vectors.

Google's AI Overviews: Major Flaws Emerge

a google sign in front of some bushes and trees

Google's new AI-powered search feature, designed to provide concise summaries, is facing scrutiny over inaccuracies and bizarre responses, raising concerns about the reliability of generative AI in search.

Why it Matters: The rise of generative AI threatens Google's dominance in search, making accuracy crucial for retaining user confidence and ad revenue.

Zoom In: Examples of AI Overviews' failures include recommending adding glue to pizza, eating rocks daily, and listing the health benefits of dangerous activities.

Numbers:

Google controls 86% of the search market.
8.5 billion daily searches on Google.
$240 billion in annual ad revenue for Google.
AI adoption is projected to reach 78 million US users by 2025.

Big Picture: The mistakes highlight AI hallucinations - when generative AI presents false information as fact due to flawed training data or algorithmic errors. Google faces scrutiny over quality control as it competes with ChatGPT and others in the generative AI era. Losing user trust could undermine its search dominance.

Reality check: While some failures involve unlikely queries, even occasional inaccuracies raise concerns about AI Overviews' reliability and Google's ability to fact-check AI-generated content effectively.

What's next: Google promises improvements, but publishers worry about losing traffic if summaries satisfy user needs. The search giant's core systems face new challenges in the generative AI landscape.

AI Investment Deals

ifeel, a Madrid-based AI mental health solution provider, raised $20M in Series B funding led by FinTLV Ventures and Korelya Capital, bringing its total funding to $40M.

Maven AGI, a Boston-based AI customer support startup, raised $20M in Series A funding. Its platform resolves 93% of customer inquiries and reduces support costs by 81%.

EthonAI, a Swiss AI manufacturing startup, raised $16.5M in Series A funding. EthonAI uses AI to identify defects and improve manufacturing processes, serving clients like Siemens and Lindt.

Bootloader Studio, a Singapore-based affective AI and spatial computing startup, raised $5M in seed funding led by Antler Elevate. The company focuses on technology that reads and responds to human emotions for two-way engagement.

Maxa, specialising in Snowflake Native Apps for business systems and ERP AI insights automation, raised $21M in Series A funding.

Zendata raised $2M, led by PayPal Ventures, to develop AI governance and data privacy solutions.

Infographic Of The Day: Most Impersonated Companies

This circle graphic shows the top companies impersonated in online scams in 2023.

Others

Perplexity AI Introduces Pages Feature for Report Generation

Perplexity AI generates comprehensive reports and articles from a single prompt, leveraging real-time web searches to provide accurate information.
Users customise and verify sources, ensuring high-quality, tailored content for various audiences, including content creators, educators, and researchers.

US Slows AI Chip Exports to Middle East Amid Security Review

The US has delayed issuing licenses for large-scale AI accelerator shipments from Nvidia, AMD, and other chipmakers to the Middle East as it conducts a national security review, concerned about potential technology diversion to China.
This move affects sales to countries like the UAE, Saudi Arabia, and Qatar.

ACLU Calls for Investigation into AI Hiring Bias by Aon Consulting

The ACLU has filed a complaint with the FTC against Aon Consulting, accusing the firm of using AI-powered hiring tools that discriminate against candidates based on disability and race despite marketing the tools as "bias-free."
Aon's ADEPT-15 personality test and other AI-infused hiring systems are alleged to disproportionately impact autistic individuals and people of colour, prompting the ACLU to demand an investigation and corrective actions from the FTC.

Apple Plans Major AI-Driven Siri Overhaul for App Functions

Apple is set to unveil a significant upgrade to Siri, allowing users to control individual app functions through advanced AI as part of its broader AI strategy.
The new AI-powered Siri will initially support Apple's apps and is expected to feature enhanced capabilities like voice memo transcriptions, automated message replies, and advanced photo editing.

Like TheAIML?

Please don’t keep it to yourself. Colleagues and friends can sign up here.

For even more,

Our newsletter archive: Click here
Follow us on Linkedin,
Follow us on 𝕏/Twitter

Your feedback is crucial in helping us improve and deliver the content you want to see. Please Email us