Probably related to Inside Facebook Reality Labs Research: The future of audio - Tech at Meta
Facebook?!! So they buy what may be the best HA technology available and say to customers, âWe have the best HAs money can buy, but you canât use them without being logged into your FB accountâ.
Buried in FBâs privacy statement is the notice that FB will be recording everything the HA hears, including everything you say and everything that is said to you, and all the conversations around you. And everything that the HAs pick up will be recorded, transcribed, archived, and conglomerated with everything you type into FB or any website that has FB affiliate links, and used to construct an incredibly detailed and comprehensive picture of you and every other FB HA user.
But hey! These are the best HAs you can buy and weâll sell them to you for $2500!
2 Likes
Wow! Fascinating read. Thank you.
Who wouldâve thought Facebook? But after reading that, it makes perfect sense. We just donât know what new technology is right around the corner, do we?
1 Like
I could be wrong, but Dwight Crow and Andrew Song are both ex-Facebook.
Youâre not wrong. They go back about 6 or 8 years since they left FB.
I still think they need to hide the brain inside of something that people need to carry with them anyway- a mobile phone. That means Apple or Google to me. Both design their own chips. Google at least already have AI processors on-board.
But couldnât Oticon or Facebook or whomever design an app where the HAs use the phoneâs processing power to achieve what the Brain did?
Way above my pay grade! Going by things I half-remember reading or hearing about⌠I think the architecture for an efficient AI core is very different to that of a general-purpose core. I have no idea how tightly coupled the architecture of an AI core is to the application thatâs using it. Would you need a âWhisperâ core or could you repurpose another? Youâd think that Whisper might have gone that route if it were possible.
Or maybe it was the need for ultra-low latency communications that lead to the brain. A few years later ultra wideband is around the corner and that might be the missing link?
What this Facebook article brings to light is just how quickly technology is evolving. There are new applications just around the corner that we havenât even imagined yet . . . But someone in Silicon Valley or some other research lab has. That was the beauty of Whisper. A subscriber would be able to keep abreast of the new technology as it was emerging without having to buy new hearing aids to be able to take advantage of it.
So, as he sit here having to face the prospect of reentering the hearing aid market, Iâm looking forward rather than backward at what might be coming onto the market rather than whatâs already there. Whisper seems to be giving us 6 months notice with a full refund of what weâve already paid into our accounts. So, that gives me a chance to see whatâs on the horizon.
New hearing aid models often seem to be announced in the fall with availability coming soon after that. Itâs been almost 3 years since Oticon announced the introduction of the more with its Deep Neural Network. Wides and Starkey followed in quick succession with their own Implementations of AI and Phonak made their aids 360. In that same span Whisper has come and gone and Philips has brought AI to Costco. Iâm sure Iâm missing something, but my point is that there are now multiple options available that didnât exist just 3 years ago.
So, how do we get the inside track in the industry on whatâs going to be announced this fall?
My $0.02:
I would guess that the limitation of AI in the stated environment, that is hearing in noise, will be a challenge (to say the least).
Most noise cancelling products do well eliminating constant frequency noises, like jet engines, tires on the highway, air conditioners, etc. No AI involved (usually) just electronics using feedback.
Like many, the noise I want most to overcome is unwanted voices in the background. Not much different than I wanted 65 years ago. So, if AI can remove all voices, how will I hear my wife? Hmmm, maybe some possibilities there. 
In my simple mind (I did programming in the past), AI would have to have âbadâ voices in a model (file) or âgoodâ voices in the model or both. Perhaps I could record the good guys in my model and let only those pass through. Then add other unwanted sounds. So we would have to funnel the incoming sounds into a process that would take enough of a clip to compare it to the sounds in the model so as to make the âdecisionâ.
We do similar with surveillance camera AI. If you want to recognize persons in a video camera stream, you have to have a model of thousands of persons in thousands of poses. So, the software looks at the first few frames and, if there is a person in the wanted zone of the image, it records an alert. Otherwise, it cancels it. Alas, although very accurate for all sorts of objects (animals, people, vehicles, etc), the time for analysis is usually over 100 milliseconds. We add the use of the graphics processor on the video card to improve significantly. Facial recognition adds a lot more.
Doing that, via Bluetooth, to a processor or smartphone could be challenging. Too much latency for live listening, I would think. Not sure. Wonder who is working on something like that?
Sounds like Whisper had the beginnings of the idea but (maybe) couldnât find the technology. Just a WAG. Maybe they could have run wires from the HAâs to the brick, er brain?
My grandfather had a âbrainâ hanging around his neck connected to a single earpiece by a large pair of wires (1948). State of the art at the time.
Oh, our surveillance camera AI still thinks the neighborâs cat, including his shadow from a night light, is a person. Angers the cat.
Edit: I should add that we have a temporary fix at our weekly koffee Klatch. When the voices get too loud to hear in the place, we take turns yelling âheyâ real loud. That provides quiet until the voices build again.
1 Like
Louyo, Whisper actually accomplished the separation of speech from background noise very well and without lag time. Thatâs why Iâm disappointed to seem leaving the arena. Of course youâre correct that separating speech from background noise is impossible when the background noise is OTHER speech.
I am disappointed as well. I think too much focus here was wasted on their business model. Part of the âyour aids wear army bootsâ attitude by a few.
My aids have a front focus program that works very well for me in noisy places. Hopefully, the next âWhisperâ will come along and we will progress.
1 Like
Intact human hearing separates speech from background speech, so itâs certainly possible.
I donât know what to think about the success, or lack thereof, of Whisperâs efforts. Now we know theyâre pivoting to a different product. Were they putting full effort into the hearing aids for the last year, or were they trying out potential new directions?
Maybe their hearing aid goal was achievable with continued work, but another product using the existing teamâs skill set offered a surer or more lucrative exit i.e. eventual buyout or IPO. Just a thought. The venture capital investors have a lot to say about the companyâs direction. And if you were an employee or executive that believed in the original mission, would you turn down a surer chance of making âeff-you moneyâ in order to stick to that mission, if you even had a choice?
Below is the letter I just got from Whisper. Weâre getting a full refund. Whisper is collecting the system just to dispose of it, so thereâs no reason not to keep it. But Iâm going to replace it with a more polished system for daily use, since my long-outstanding Whisper issues (no media streaming, flaky Bluetooth, music program not separately tunable) arenât getting fixed.
========================================================================
Thank you for being a loyal Whisper customer. We have enjoyed helping people hear better, but have made the decision to complete a voluntary market withdrawal of our Hearing Systems as our business direction changes, effective June 8th, 2023. Donât worry, there are no known safety concerns with the devices, but this means we will no longer service, produce, or support the Whisper Hearing System.
We hope you quickly find a new replacement hearing device that meets your needs, and to minimize any inconvenience, we are providing you a full refund for your Whisper Hearing System.
Should you have any questions, we will continue to provide limited customer service support to facilitate product returns and refunds until December 1, 2023, after which we will no longer offer support for your Hearing System.
- We ask you to return your Hearing System to Whisper as soon as possible so that we can properly dispose of it but no later than December 1, 2023.
- You are eligible for a refund whether or not you return your Hearing System to Whisper.
How to Make a Return
Within the next two weeks, you will receive a return shipment box and pre-paid label for you to ship your Hearing System back to us for disposal. You can drop off the box at any FedEx location. If your shipping address has changed, please contact Whisper customer service at 844-936-4327 or by emailing support@whisper.ai to update your address to ensure you receive the return box.
System Refund Information
As an Allegro Credit financing customer, no future payments will be processed on your account and you will be provided a full refund for any payments made, including tax and service fees. This refund will be issued by Allegro Credit to your listed payment method within the next 3 to 6 weeks and any outstanding payment balances will be paid on your behalf.
You are eligible for a refund whether or not you return your Hearing System to Whisper.
If you have any questions, please contact us at support@whisper.ai or give us a call at 844-936-4327.
Thanks again for being part of our journey.
Andrew Song,
CEO
Whisper
1 Like
Well you really canât ask for more than that. Full refund of everything you ever paid. Return the HAs at the companyâs expense as soon as you replace them. Or keep the HAs and your audi can use Whisperâs software to adjust the HAs until December, and you STILL get the full refund of everything you have ever paid.
The next big thing in both mobile devices and desktop computers is the Neural Processing Unit (NPU). You have your CPU, your GPU, and now itâs going to be any device worth having will have an NPU. Apple and Google already incorporate such devices into their smartphones. But supposedly the success of OpenAIâs ChatGPT is going to drive further upgrades to the smarts that come with your devices. There is a rumor that to upgrade to Windows 12, the best experience will be with computers having NPUs if you want fast results.
Perhaps if any market is big enough to be worth addressing, HA OEMs will write âbrainsâ specific to a type of device. So for AI, the âMFi AI deviceâ approach again might put Android at a disadvantage if every Android OEM plays cute with a different sort of NPU in its mobile device whereas Apple strikes a deal with HA OEMs, âHereâs our iPhone NPU. If you want your HA brain to work gangbusters with our device, use our NPU API to develop your brain firmware, etc.â
It will be interesting to see how Appleâs privacy selling point works against edge computing. The most powerful combination would be to process locally but train and upgrade the device through sharing speech data to the cloud. If the data never leaves your device, Apple might be at a disadvantage to Google and Microsoft, who ask users to share data to the cloud (usually in a nonidentifiable way) to improve the software process. Perhaps in a difficult listening situation, the cloud could rapidly learn from data submitted how to improve the local algorithm in an hour or two spent in a particular noisy restaurant, as the available cloud resources could be far more powerful than anything available on a local device. BTW, HA OEMs, at least ReSound, already ask the user permission to share some HA data back to them to monitor and improve their devices. I always agree to do so.
I will be keeping my Whisper hearing aids because they do everything I want in terms of enabling me to hear better. I donât care about blue tooth or any of the rest of the other ancillary functions. Because of a glitch in one of my hearing aids, they replaced my entire system free of charge just 2 months ago, so I essentially have brand new hearing aids . . . for free. Every major hearing company upgraded or replaced their platform within the past year, a couple of them as recently as this past Feb/Mar. At a minimum my Whispers will get me through the next couple of years until the next generation of new upgrades. Thank you, Whisper.
2 Likes
I donât see why any Whisper owner wouldnât do what you plan to do, Bill. Thereâs nothing further to gain by returning your system back to Whisper because you can get a full refund regardless, so it would actually be a dumb thing to do to return the system back to Whisper.
If one doesnât want to continue to use it after December because of lack of support, it can still be relegated as a backup system for free of charge. But if not actively used anymore and only relegated for backup purposes, I would still make sure that I donât leave the brainâs battery fully depleted or fully charged while in storage because that will kill the battery much faster.
1 Like
I canât help but wonder if releasing a hearing aid product out to the market was just a âplayâ in Whisperâs playbook for its ulterior strategy to provide a proof of concept product out to the market, as a stepping stone in qualifying for a buyout by a big tech company that it might even had had in mind already from day one. This is simply because they seem well prepared and well funded to give a full refund to customers for the trouble of using their product â the cost of executing a strategy. Itâs not really a sign of a troubled startup running out of funding and forced to go out of business.
Their very unique product model (with the brain) and financial model (the lease) in the HA industry sure made them stand out and help catch attention easily. The long term viability of their product model and financial model was never a real concern (like I thought it would have been) if it were just part of a strategy to catch attention and get sold eventually, and soon enough.
I wonder if they really were working on a new hardware platform and newer and more frequent functionality software upgrades like they said they were, or whether they were more busy trying to get bought out before they had to deliver any of the new stuff.
1 Like