Exciting news! Be My Eyes is going live with Be My AI

I can’t say I’ve ever thought of doing a podcast. I’ll give it some thought. I don’t have a good microphone though. I think I’ll sit this one out and maybe give it a go in the future. As for response times. It’s between 3 and 8 seconds on average I’d say it’s usually toward the lower end but can occasionally take quite a while. It’s worth the wait though and BME are definitely trying to get the time down as low as possible. I think it depends much less on the phone you’re using than you might imagine. During beta testing I saw it reliably at 2 - 3 seconds but the more they try to make the model give more useful information the slower it will be. I think things will speed up over the next year or so as more GPU’s are manufactured and added to OPEN AI’s data centres.

If you would like to email me, I can put you in touch with Clay at Be My Eyes. He’s their legal person. It might make sense to join forces. I’d love to read the paper either way. andylanesemail@gmail.com

Android beta is running at the moment if you want to register for it. I’m sure they’d appreciate your feedback.

At the moment, It’s just a pleasant tune but we’ve already submitted feedback that some might prefer haptic feedback. I know development will have been slowed down a bit while the face processing issue was resolved but hopefully updates will start coming in thick and fast. It’s in the queue though.

I believe they would prefer to keep it closed for the time being at least. They would like as much feedback as people want to provide but managing 10’s of thousands of people in a slack channel would be a full time job for someone and I don’t think they’ve got the manpower to spare on that. It’s all hands to the pumps trying to make Be My AI as good as it can be. They are also working on lots of other advocacy projects which we will all benefit from. If you have feedback, send it to them through the email address on their website. I know they value ideas from the community and want to hear about bugs people are experiencing.

Had to get some help on this one as I’m not an android user but here’s how to do it.
You have to make sure your Google play, the device, and Be My Eyes are all signed into the same account. Then you have to search for Be My Eyes in the Google play store and there’s an option in there that says get the beta or join the beta or something like that. Once you click that it takes a few minutes for it to add you to the beta program.

I hope that works.

Yep you’ll occasionally get some strange descriptions but not too often. You’ll also get some incorrect information presented with absolute confidence sometimes. These are known as hallucinations which I’m not keen on but thats another subject. For this reason it’s not a good idea to use Be My AI for something that could cause danger if it was described incorrectly, medicine is a good example. Being as accurate with the camera as possible will definitely help and try to keep your hand as still as possible when you take the photo. Interestingly, you can get Be My AI to help you with your photography skills if you’d like. You can ask it questions about how to make the photo better so it gets a clearer view of what you’re trying to capture. It’s usually pretty good at helping you out.

We have already requested a way to export photo’s from Be My AI with the description already attached as alt text. This should make your job a little easier. I don’t know when it will be or if it will be included but fingers crossed.

Be My Eyes are doing a lot of advocacy work for the blind community in bringing this product to market there are hurdles they’ve had to jump over. I also know of advocacy they’re doing around accessibility standards on the web. With your legal qualifications I’d imagine they’d be interested, Everyone talks about lived experience and its great that is becoming more valued but having someone with lived experience and subject expertise is hopefully even more valuable.

Sadly I don’t think Be My Eyes are at the point of mining historical interactions for future improvements but I wouldn’t be surprised to see it happen in the mid term. For now though I’d send feedback to Be My Eyes directly. If you did want to correct Be My AI, I reckon that will go into a data set at some point in the future though, I’ve been doing it myself. I just know it’s not being used yet.

Is it still happening? Mine is working.

Way back in the beta cycle I suggested a few tweaks to this screen to help data mining but they aren’t really doing that yet. All they are doing at the moment is looking at the I had a problem records to see how they can improve Be My AI. A sophisticated data mining records using AI may be something they do eventually but theres a lot of other things they want to get to first and this would be a massive project. The potential of mining that data is vast especially when an AI can be self improving but not at the top of the development hopper.

As I’ve never had this issue come up I can’t offer anything more than the generic restart your phone first then delete the app and re-install as potential solutions to this problem. I’d suggest getting in touch with Be My Eyes if the generic advice doesn’t work. There may be something specific to your account they can see but I’d go with the generic stuff first as it usually fixes most problems.

If you were sending all those pictures in the same thread I may be able to help you speed things up. Each time you take a new photo and or send another message it actually uploads the entire thread of messages and photo’s with each new photo or message you send. So you aren’t actually sending 1 photo or question, you’re sending the entire thread, every time. This is why it will get slower and slower the longer the thread gets. It has to be this way so Open AI can see the entire thread so the AI can answer your most recent request while keeping the rest of them in mind as they might still be relevant. Just like a conversation between 2 people. If each new request forgot the previous one, we’d be in a bit of a mess. The solution however is really simple. Once you know you don’t need to refer back to anything that has already passed or the thread is getting too slow. Just clear and start fresh. You’ll get the faster responses back again. I imagine in time, Open AI will be able to hold that thread on their servers so you can just send your most recent question or photo and it will still see the rest without uploading each time but that hasn’t happened yet. It’s the same on the Chat GPT app and Bing. They all upload everything every time.

Hey, you shouldn’t be having that problem even if you are in Illinois. Try closing the app, even restarting your phone. All the usual stuff, could even try re-installing the app if you’d like but if that doesn’t work then send a quick email to support. They are really helpful. I did have 1 picture fail earlier. Possibly because of increased demand with all the new people joining. It’s an enormous expansion in a very short period of time so the odd failure is understandable until they learn how to cope with the enormous demand and iron out the bugs. It shouldn’t be failing repeatedly though so definitely something else going on.

Your bio says you are on mars and I don’t think they have passed any outdated legislation on Mars that should prohibit Be My AI so you should be okay. Haha.

Its always a good idea to clear the chat every now and then, maybe every 5 - 10 messages unless you need it to know about something further up the thread to be able to answer question. The reason is that every time you send a message or photo, it uploads the entire thread, every time. So if your thread is 50 messages long with 20 photo’s and you send a photo, it has to upload 26 photo’s and 50 messages. This is so that it understands the thread and can make the answer relavant to anything you’ve asked for earlier on in the thread. Open AI don’t yet hold the thread on their servers. With Chat GPT it’s the same but you don’t notice the delay because it’s all just text. with BMAI it obviously includes photo’s which take a lot more time to upload. I just refresh the thread every 5 messages or so. it really speeds it up and cuts down on the error messages.

To my knowledge this is being actively worked on however I wouldn’t expect a solution next Wednesday. Realistically this is going to take some time. What I can tell you is that everyone who would be involved in making this happen, wants to make this happen. There is nobody saying no however it’s a very sensitive issue that is going to require careful consideration and is likely to take some time. What should be allowed, what is absolutely not okay under any circumstances. These are most definitely not easy questions to answer and will take some serious efforts. Nothing like this has ever been possible before so it’s all new ground people are treading on. My personal opinion is, we’ve got something incredible right now and all the right people are working on making it everything we could ever hope it would be this includes developments ranging from adult content to full video support and much much more. Sadly Rome wasn’t built in a day but my god have we got some great builders on the case.

Just to preempt the comment that everything should be allowed and how dare they choose whats okay. If you have a think about it for a minute. i’m confident we can all come up with things we’d never want described in a million years. Things that are repellant illegal and we’d all rather didn’t exist in the world. Part of the problem is going to be ensuring that type of situation can’t happen. This is a very complicated problem to solve which is why I wouldn’t expect it to happen quickly. Patience will be needed on this one.

This has been an ongoing conversation since the early days and the result is something along the lines of whats the point in describing something to a blind person if you can’t describe reality. If I ask if my hair is dirty and it says no your hair looks perfectly clean when actually it looks filthy then theres no point in asking. If I’m trying to get ready to go to work and I ask if I look good and it says you look wonderful when I’ve actually got a stain on my shirt then again not so good. There are literally endless situations we can think of where not wanting to cause offence would be the easy route to take but completely useless as a tool to describe the world to blind people. I think we’re going to all need to put on our big boy pants for this one and recognise that reality can be offensive but having reality described is the point of the app. It’s a tool to access a visual world and so should be accurate not political.