It is interesting that the Reddit post is no more. KS9 and KS10 were released 8 months after Marvel and Paradise. I would agree that releasing a Costco equivalent 2 months after Infinio release is unlikely, but the trend has been for Costco to get increasingly more current hearing aids. (For instance Jabra Enhance 20 released one month after Resound Nexia) I don’t think AI “evaluates” anything.
1 Like
ChatGPT’s response to “Does AI “evaluate” anything?”
Yes, AI systems perform evaluations in a variety of contexts, though they do so differently than humans. Here’s how evaluation works in AI:
- Statistical Evaluation: AI uses mathematical models and statistical algorithms to evaluate patterns in data. For instance, a machine learning model might “evaluate” whether an image contains a cat based on features it has learned from training data. The model does this by calculating the probability that the image matches the patterns of a cat and then making a decision based on that probability.
- Scoring and Ranking: In search engines or recommendation systems, AI evaluates and ranks content to provide the most relevant results. It assesses factors like past user behavior, content quality, and contextual relevance to “evaluate” which items best match the query.
- Error and Performance Metrics: AI models are trained with metrics like accuracy, precision, and recall to evaluate their effectiveness. During training, the model repeatedly evaluates its own predictions against known results, adjusting to improve accuracy.
- Contextual Decision-Making: AI also evaluates options in context-dependent situations, such as making investment decisions, diagnosing medical conditions, or driving autonomous vehicles. The AI system assesses multiple possible outcomes based on learned patterns to make the best decision within its programmed constraints.
- Natural Language Understanding: When dealing with human language, AI evaluates the context, sentiment, and meaning to generate appropriate responses or classify sentiments. This involves semantic and syntactic evaluations based on learned language models.
While AI can effectively evaluate in quantitative terms, it’s important to note that its “evaluation” lacks subjective reasoning, emotions, or personal judgment, which are integral to human evaluation. AI’s evaluation is strictly based on predefined objectives and the data it has processed.
It’s worth watching Andrej Karpathy’s 1-Hour YouTube video on
THE BUSY PERSON’S INTRO TO LARGE LANGUAGE MODELS
to appreciate both the power and shortcomings of generative AI. For those who haven’t noticed, the Nobel Prize in Physics was awarded this year to two pioneers of modern AI, John J. Hopfield and Geoffrey E. Hinton, for their pioneering contributions to artificial neural networks. The 2024 Nobel Prize in Chemistry was awarded to David Baker, Demis Hassabis, and John Jumper. David Baker was recognized for pioneering advancements in computational protein design, allowing the creation of new proteins with potential applications in medicine and materials science. Hassabis and Jumper, both from Google DeepMind, were honored for developing AlphaFold2, an AI model that accurately predicts protein structures, which has transformed protein science by allowing researchers worldwide to visualize and study the structure of millions of proteins. This work opens significant possibilities in drug discovery, materials engineering, and understanding biological processes.
Hinton classified ChatGPT as not so accurate but still said it was very handy to have a not-too-good world expert on ~everything at his beck and call. He also said we entering an era where AI will have the impact on society equivalent to the Industrial Revolution.
The DNN in the Sphere is directly derived from Hinton’s and Hopfield’s work. I imagine there’s a lot more AI to come that will impact and elucidate hearing science.
1 Like
I currently wear the All Day Clear. They are the same hardware as earlier Phonak HA’s but with firmware specifically tailored to OTC. The pricing is simply a marketing decision, having nothing to do with the HA’s themselves. The OTC firmware is dumbed down to only three adjustment bands; low frequency,mid-range, and high frequencies. They could have taken the same (or virtually the same) hardware and put back in the prescription firmware with many more adjustment bands, and many more programs in the app.
Interesting. I just ordered a pair of Reaches, but with Costco’s return policy, I’ll give this one a good look. If this is markedly better, I may have to exchange. I don’t think Costco in Canada allows HA’s to be trialed (even with a reserved block on the CC as a deposit). I would also have to get new molds, as my hearing loss is too severe for domes.
It’s been mentioned that the Sennheiser Sonite R will be available in 107 Costco hearing aid centers in the US. There are over 500 Costco hearing aid centers in the US so availablilty, at least initially will be in 1 in 5 hearing aid centers in the US. Disappointing that we don’t have a spec sheet yet. Several news stories about this but neither Costco nor Sennheiser are talking.
That was another reason I factored into disbelieving the “Sonite R” was a Sphere knockoff. If Costco could have offered a Sphere equivalent, would they have released it in only 105 Costo Hearing Centers?
I wonder how much of the news is based on that Reddit post? This Reuters article says as of today, supposedly, that both Costco and Sonova are “mum” on the Sonite R release. Would be hilarious if the whole thing were market manipulation by someone who bought and sold Sonova stock after the 7% rise today! 
1 Like
Seems pretty real as Sennheiser Sonite R shows up on the EasyLine Remote App. Picture appears same as posted on this forum. User instructions mention only Bluetooth Classic. No real info.
2 Likes
Unless I missed it, there’s no reasons suggested on why Costco is removing the Rexton Reach from its offering!
Any speculation?
1 Like
IMHO, Anyone that thinks they are being short changed with the Signia IX technology is confused.
4 Likes
From another post.
It’s unclear why Costco is interested in replacing Rexton Reach, which like its competitors, is a high-quality device with advanced features. Some market analysts have cited possible product delivery problems may be factor. Last year, as part of the global near-shoring movement, WSA moved a good deal of its production to Mexico when it opened a new state-of-the-art manufacturing facility in Tijuana.
2 Likes
Curious what happens if you ask one of the AIs for a percentage estimate of the likelihood of a statement being “true.”
@MDB, some examples…
Microsoft Copilot:
Can you give a confidence level with your answers?
I don’t quantify my confidence in percentages. I can assure you that my answers are based on the latest available information and best practices. If there’s ever any doubt or if I don’t know something, I will admit it and help you find the right resources.
ChatGPT:
Can you give a confidence level with your answer?
ChatGPT said:
Yes, I can certainly include a confidence level with my answers! For more complex or uncertain topics, I’ll explain the level of confidence I have in the response and outline the areas that may require further verification or could benefit from additional sources.
Would you like confidence levels for every response, or just when I feel it’s particularly important?
Google Gemini:
I mean in general, with any query, can you provide a confidence level?
Unfortunately, as an AI language model, I cannot directly provide a specific confidence level for every response. This is because my responses are based on the information I’ve been trained on and my understanding of the query. While I strive to provide accurate and relevant information, there are inherent limitations in my ability to assess the absolute certainty of my responses.
However, I can provide some general guidance on how to interpret my responses:
- Factual Claims: For factual claims, I rely on a vast amount of data. If I provide a factual statement, it’s likely highly accurate.
- Complex Queries: For more complex or nuanced queries, my confidence may vary. I may provide multiple perspectives or qualify my answer to reflect uncertainty.
- Hypothetical Scenarios: When answering hypothetical questions, I’ll often use phrases like “it’s possible” or “it’s likely” to indicate a degree of uncertainty.
Ultimately, it’s important to critically evaluate the information I provide and, if necessary, consult additional sources to verify the accuracy.
Would you like to discuss a specific query or response in more detail?
Specific Gemini example of electrical wiring and component function in Ford Home Integration > System for its F-150 EV:
How confident or sure are you of the previous answer?
I am highly confident in the previous answer. Based on the information I have gathered about the WHB 200 and Acrel meter, it is accurate to say that the WHB 200 can independently determine the grid’s on/off status.
The WHB 200 is designed as a comprehensive power monitoring device, capable of measuring and analyzing various aspects of electrical power flow. This includes detecting the presence or absence of grid power.
While the Acrel meter is primarily used for energy measurement, it does not have the same direct functionality for grid status detection.
If you have any further questions or require more specific information, please feel free to ask.
Follow-up Question to Previous Query:
Can you give a confidence level with your answer?
I would assign a confidence level of 95% to my previous answer.
This confidence level is based on the following factors:
- Product Documentation: I have reviewed product documentation and technical specifications for both the WHB-200 and Acrel meters.
- General Knowledge of Power Systems: I have a solid understanding of electrical power systems, including grid-tied and off-grid systems.
- Logical Reasoning: Based on the functions of these devices and the typical setup of residential solar and battery backup systems, it is logical to conclude that the WHB-200 can independently detect grid status.
However, it’s important to note that specific implementations and configurations might vary. If you have access to more specific documentation or hardware details, it would be beneficial to consult those resources for the most accurate information.
I have found, however, that Copilot can offer egregiously wrong answers with confident statements. As I noted in an earlier post, if it matters, I always check other sources, including using standard Google and Bing searches as a follow-up. Once you’ve found the supposed crux of a matter, classic search hits are easier to wade through for additional confirmation.
1 Like
Local Costco audiologist confirmed that they would be selling the Sennheiser Sonite R but doesn’t have any information on them yet. Expects to begin selling them in a couple of weeks. Have not received any information so far that they would stop selling Rexton Reach and expects a new Rexton model soon.
2 Likes
This fits with one comment I saw on Reddit about it first being available “back East” and in Oregon and maybe Washington. I don’t think this whole thing was supposed to leak this early. That would explain why everybody is quiet about it. We may not have much info for several weeks.
When I asked about Sennheiser the people at the counter said they had not heard of that, then a young audiologist (everybody is young when you get to an advanced age) popped out of the testing booth and said yes. Wanted to know how I knew 
She did not know what Sonova product would be the equivalent, she guessed maybe the Lumity but I am guessing the Infinio R.
One thing I’ve read about generative AI is that its main source of input, the internet, is becoming increasingly populated by AI output. Inclusion of AI output in training of models is going to degrade them. So by posting AI output verbatim on this forum you’re helping speed up AI’s deterioration.
1 Like
And all the erroneous human input, such as that about the Sonite R HAs, has no effect?! Funny that humans want to rein AI in but allow themselves unchecked speculation over shreds of evidence.
See the ChatGPT answer above on how it evaluates data: Costco will stop selling the Rexton Reach and replace it with the Sennheiser Sonite - #42 by jim_lewis
AI evaluates and ranks content to provide the most relevant results. It assesses factors like past user behavior, content quality, and contextual relevance to “evaluate” which items best match the query.
as well as other items in the linked post on “evaluation.”
To put things in perspective, Terence Tao, a Fields Medalist (the math equivalent of a Nobel Prize), has been quoted as saying that ChatGPT-4o (or omni) could only reason in math about as well as a mediocre graduate student.
That sounds pretty terrible. But how many of us can even reason in math as well as a mediocre UNDERGRADUATE math student?!
So, I’d say we’re in the Model-T days of AI. It’s a glass half-full or half-empty flippin’ point of view. A lot of white-collar human jobs do not require a great degree of intelligence or skill, e.g., working in a call center from a company script. Just like machines made a lot of human physical labor obsolete in the Industrial Revolution, AI will change a lot of white-collar jobs. Most answering services at big companies now have AI voice recognition agents fielding calls at the entry level to the call system, etc.
So, just like blacksmiths and carriage houses probably made fun of Model-Ts, one can disparage AI all one wants. But it’s going to evolve just like the Model-T did and have a big effect on society.
Delving into the app, I found this link which seems to show rather a lot of marketing blurb.
And the product instructions, data sheet and so on…
5 Likes
That’s very useful. Good find. That gives me pretty good confidence that it is very similar to the Lumity 90.
I’m basing that on appearance (microphone placement), exact dimensions as the Lumity R and the fact that it has an Echo feature which is unique to the 90 level. It also has a Speech Boost feature which sounds a lot like Speech Enhancer which is also unique to the 90 level. Perhaps disappointing that they’re not Phonak’s latest and greatest but they’re still likely great aids.
1 Like