The cocktail party problem

During a hunting event, Napoleon’s staff released hundreds of rabbits for him to hunt. Instead of running, they swarmed him. Bunnies, unite!

What’s in store:

• Finding your fave photos just got effortless with Google’s “Ask Photos”.

• No more “What did you say?” moments, AI solves sound issues.

• Around the web: Apple’s upcoming Watch Series 10 may feature a sensor that can detect sleep apnea.

• Learn AI skills for free with Growth School!

Read Time: 6 minutes

AI BREAKTHROUGH

AI solved an age-old problem… about parties

Ever tried having a deep convo at a house party with music blasting? It can be surprisingly hard, until you “focus in” on the person you’re talking to, at which point all the background noise seemingly disappears.

The "cocktail party problem" is a common issue we’ve all experienced: trying to hear one person in a noisy room.

While humans manage this in person, technology adds a layer of difficulty, making it much harder to distinguish who is speaking on a recording of the same circumstances.

This has had some pretty serious consequences - making audio evidence far less useful in court, for example.

Recordings with multiple voices can make it hard to identify who’s speaking, making the evidence unusable.

Keith McElveen, founder of Wave Sciences, faced this problem during a war crimes investigation.

Separating voices from one another was a major challenge.

His solution came through… you guessed it, AI, which can help locate and filter out sounds based on where they come from in a room.

This technology also considers how sound reflects off surfaces, which complicates things further.

McElveen's AI "backtracks" sound to its source, much like a camera focusing on a subject while blurring the background.

Although the audio might not be crystal clear, it’s accurate enough to be used in court cases, like murder investigations.

Since then, the technology has passed multiple tests and is now being marketed for use in areas such as sonar analysis, smart speakers, and hearing aids.

Here’s how AI helps:

• AI can isolate voices in recordings, making previously unusable audio evidence valuable in court.

• The technology mimics how human hearing works, solving a longstanding acoustic problem.

• It’s being used in forensics, sonar analysis, and voice interfaces like smart speakers.

Listening like a human but better

Wave Sciences initially used a large number of microphones to separate voices but soon realised this was not practical.

After ten years of research, McElveen developed an AI solution in 2019 that uses fewer microphones.

The AI tracks sound in a room, identifying its source and blocking out any sound that couldn’t have come from the speaker.

In its first real case, the technology helped in a US murder trial by turning unclear recordings into crucial evidence.

The FBI used it to record conversations in noisy restaurants, making previously inadmissible audio clear enough for court.

The tech is also being applied by the US military for sonar analysis and could be useful in situations like hostage negotiations and virtual reality.

AI's role in audio analysis is expanding rapidly.

Machine learning models are now used in criminal investigations to identify speakers and authenticate voice recordings.

Companies like Bosch are also using AI to predict machine failures by analysing sound.

McElveen even suggests that the AI’s methods may be similar to how the human brain processes sound, potentially solving the cocktail party problem in the same way our brains do.

So, soon I won’t have to pretend I heard what someone said in a noisy room?

Vote for live results and see results + opinions from yesterday at the bottom of the email.

LEARN AI SKILLS

🦾 Master AI, ChatGPT and 20+ AI Tools in just 3 hours

Don’t pay for bad AI courses when you can learn AI for FREE!

This incredible 3-hour Workshop on AI & ChatGPT (worth $399) makes you a master of 25+ AI tools, hacks & prompting techniques to save 16 hours/week and do more with your time.

Sign up now (FREE for First 100 People) 🎁

This workshop will teach you how to:

• Do AI-driven data analysis to make quick business decisions

• Make stunning PPTs & write content for emails, socials & more in minutes

• Build AI assistants & custom bots in minutes

• Solve complex problems, research 10x faster & make your simpler & easier

You’ll wish you knew about this FREE AI masterclass sooner 😉

On This Day

September 8th 1966 - "Star Trek: The Original Series" debuts with its first aired episode, "The Man Trap," on NBC-TV in the United States.

AI Image Or Real? The Results

Looks like a desktop wallpaper. But was it real?

59% said AI

41% said it was real

IT WAS AI! Did you get this one?

Free Resources

Still wasting time on PPT? Well, we gathered a list of tools that could help you take your presentation to the next level without the hassle.

Download your high-quality copy here! →

GOOGLE

No more endless scrolling with Google’s AI

Google is testing a new “Ask Photos” feature, allowing users to explore their photo library in fresh ways.

First previewed in May, this feature is now available to select Google Labs users in the US.

It lets users ask questions like, “Where did we camp last time in Yosemite?” or “What did we eat at the hotel in Stanley?”

Powered by Google’s Gemini AI, the app finds and pulls up relevant photos in response to your question.

It can also summarise recent holidays or help choose the best family photos for an album. You can sign up for the waitlist on Google’s website.

Google also offers an improved “classic search,” where you can use natural language to search for images.

You can sort results by date or relevance.

This feature is rolling out in English on Android and iOS, with more languages coming soon.

Here’s what you should know:

• “Ask Photos” uses AI to organise and find images based on natural language prompts.

• Classic search is now enhanced with natural language support.

• Google Photos replaced the Library tab with a new Collection page for easier photo management.

Library tab who?

To support these updates, Google has replaced the Library tab with a new Collection page, making it easier to manage your photos and videos.

Basically, we don’t have to squint at blurry thumbnails anymore. WIN.

Mindstream Picks

Crypto: Vitalik Buterin pledged to donate all his Layer 2 tokens to charity, following Ether sale accusations.

Space: China's reusable spaceplane completed its third orbital mission, returning to Earth on Thursday, with few details released by state media.

Games: Test Drive Unlimited Solar Crown launched for $90 Gold Edition buyers, but server issues have made the game mostly unplayable.

Music: Indie band English Teacher won the Mercury Prize for their debut album This Could Be Texas, praised for its originality.

Don’t Miss - Apple's upcoming Watch Series 10 is rumoured to feature an upgraded ECG sensor capable of detecting sleep apnea, along with larger displays, a thinner case, and improved water resistance. Despite these enhancements, it may lack a blood oxygen sensor due to a patent dispute with Masimo.

AI Art

Our Image of the Day

Daily Image Prompt

Submit your artwork to Mindstream →

Yesterday’s Poll

“Is Microsoft’s plan to fix AI bias with diversity going to work?”

It’s a start, but more is needed. - 58% ✅

Yes, we need more diversity in tech. - 42%

Your Views:

“I am pleased that Microsoft is working to address AI bias with diversity, but I think the company's plan is a start but insufficient. Attempting to ensure diversity is akin to making affirmative action work—it is not a natural process. For example, we cannot simply exclude white people from advertising, as Google did for a time. However, if the machine is trained on diverse data, its output should be diverse and inclusive. This is a complex issue, and I look forward to seeing how Microsoft addresses it.” - danaleonard

“Bias is an inherent element of all cultures. You cannot weed out all bias.” - carl.d

Submit your opinions in our daily poll to be featured!