I agree with your assessment mostly, though (whether or not SSi could have afforded to do anything else) this AI business has already gone some way towards making translators and recording artists obsolete…
I know we like to be positive here, but it does bother me.
Anyway.
Thanks for the link. At least in my browser, it takes me to Spanish AutoMagic, not mp3 files. I’ve gone round in circles a few times now - I go to Challenges, it’s Welsh. I change language to Spanish… and it’s AutoMagic. No Spanish mp3 files to be seen. 
I’m on the Welsh course anyway, I just went looking out of interest to compare how much material there is for each language, and it’s weird how for me, no Spanish “Challenges” seem to exist. Not important right now though.
Well to satisfy your curiosity: I can see 2 levels of “new course” (25 lessons each, these are what I did), the tourist course (looks very handy, can’t remember if I did it), and 1 level of “old course”, which I had never noticed before.
Diolch @Hendrik - really appreciate that - we’ll definitely take you up on it in the not-too-distant future, German is very much towards the top of the list 
I certainly agree with everyone that AI is a complicated field and creates plenty of opportunity for poor behaviour, and will need careful regulation. My hunch is that over the next few years, AI will lead to increasing demands for UBI to avoid social unrest, and I hope very much we get to UBI.
We’re a very narrow usecase as far as AI goes - we’re making use of AI-enabled translation technology and AI-enabled voice creation technology to provide content, and we’re then bringing that content into our own proprietary structures in order to create courses - those structures make no further use of AI at the moment. It’s true that those technologies are going to put some people out of work (or dramatically change the nature of the work they do) - in our particular case, it’s also true that we wouldn’t have been able to afford to build more than a tiny handful of new languages otherwise.
We will (probably next year) be able to use our own datasets to start comparing learner behaviour/production which will allow us to personalise the content better and also to give insights into levels of achievement, both of which will be very high value for the individual learner - that will involve some level of closed AI in terms of how we compare inside the dataset, but that’s really just a fancy way of talking about using mathematics!
Verity, I’ll ask Ivan about accessing the Spanish mp3 files - they should be at https://en.saysomethingin.com/spanish/level1/challenge1 (as already posted) and that opens properly for me, but our membership website does sometimes do some strange stuff based on user access settings, so it’s probably our fault! I’ll let you know as soon as I hear from him
I agree with you hear too, but as an optimist I think SSi growing could create new jobs too. I think if more content is added beyond where the levels currently end then as you get further and further, the content will need to diverge more and more. That creates openings for course creators and cultural experts to guide the work, and more proof-readers to check it.
I also wonder if, in the long term, the more popular courses could be re-done with real voices to get some more variety and variation in there? But I’d rather see the effort going into the courses AI just can’t do before that.
We can keep dreaming, Aran. The exciting news that we’ve digressed a bit from is evidence that your dreams in particular can come true!
Yes, I think there are some interesting decision points ahead of us - in the short term, we’ll be putting increases in turnover back into tech dev and marketing, but if we do get through to genuine surplus, the balance between ‘real-ifying’ AI output and supporting smaller languages that don’t have the AI output will be an interesting one. I’ve certainly heard at least some AI voices (in several languages) that I wouldn’t be able to tell from humans, but if we get feedback/data that some courses are unpopular because they fall into the uncanny valley, we’ll certainly pay attention.
Perhaps the most likely outcome will be around licensing ‘real’ voices to model for the AI, once we get a bit closer to a consensus on the right balance between work/reward in that field…
Verity, I’m told that link should be fixed for you in the fairly near future
Hi @verity-davey - give it another try now. I’ve just twiddled your settings and it works for me under your account now.
Yes, that all works now, thanks. I know some people in Spain so later on I may well want these. I go offline mostly when I’m out of the house, rather than have a huge data plan, so it’s always good to have mp3s. 
It certainly sounds like you’re doing everything right to minimise any negative impact of AI. That’s great to know.
The real check is what native speakers think of the voices, of course. My phone’s satnav sounds reasonably human, but often it says things with weird intonation or the wrong emphasis. Imagine a whole load of immigrants who’ve done absolutely brilliantly learning English, but from that. It would definitely be weird. 
Of course, my phone is several years old and the satnav was made with very different tech from what’s available now!
I am really looking forward to sampling the new courses when they arrive. I have kinda caught the language-learning bug. My problem is going to be choosing one at a time and not getting so distracted by each new shiny that I end up being able to say a handful of basic sentences in several languages, but have real conversations in only two. (“Only” she says! SSiW is seriously cool for getting me this far with Cymraeg already!)
Aran, you need a new nick-name. Something based around “parhau” and “datrys”…
I think I’m going to struggle with the same challenge - already planning to do some intensive on Italian and Mandarin (and Arabic, as soon as we solve a little signalling issue for gender).
I very much agree on voices - the quality we’ve heard so far has been extraordinary, miles ahead of things like Siri 
Gary! Lovely to ‘see’ you, hope you and yours are keeping well 
On the subject of languages that don’t use a Latin-based alphabet, will / how will the words show on the screen? I would think, especially given all SSi has always said about the benefits of speaking and listening, that transliterations shouldn’t be featured… but obviously a lot of people would find it hard to believe, even with explanation, that it’s really not helpful, and it might put them off?
Showing the language written in its own alphabet probably wouldn’t have a negative effect on pronunciation, especially for people totally new to the language who can’t read it anyway… but it might be appreciated by more advanced learners who are using the course to improve?
Btw, I prefer using the mp3s anyway (helps to be able to use them offline) so it’s not a big deal for me right now, but I could never get rid of the words when I tried AutoMagic. There was a switch in the settings labelled show/hide words or something like, but it had no effect. It is so hard not to “just peek, this once” if you aren’t quite sure what you heard.
My sister is learning Japanese. She’s determined not to learn bad pronunciation with romanisations, and has started with learning to read hiragana - ie, she can sound things out, but has not yet learned actual words. Having heard me go on (and on, and on) about how well the SSi method works for me, she is interested, though hesitant about the whole not reading thing.
Obviously right now you have no firm dates, but have you a rough idea of where in the process you’ll be able to say with confidence “everything is working as it should, we can now estimate a time-line for languages, and they’ll come out in this order:…”?
Helo, I have the same problem as @verity-davey - I can only access Automagic in Spanish. Any chance you could - ahem - twiddle my settings as well? Diolch yn fawr!
Anything that helps the people of the world communicate effectively has to be a good thing, of course, but I share a lot of the reservations about AI/LLMs that other of the good folk here have raised. I just hope that Aran and SSi aren’t going to be the proverbial turkeys voting for Christmas - I’m not being rude here, I just mean, how long do you think Google will stay out of trying to out-do SSi at SSI-ing?
Full disclosure, I view Google as marginally less palatable than AI, so I’ll freely admit that my viewpoint is a tad biased, but having built a (frankly) brilliant language teaching method (and I can’t thank you enough for the few Welsh conversations it has to date empowered me to have) it would be a crying shame to see it ousted by Big-Tech.
Live performances that need real humans, in real places, at real times are about the only place that AI cannot compete (well, at least not yet!) but otherwise, as many freelance writer friends have found, slow and costly people are rather too easily expendible!
Anyway, it may never actually become a problem; long before AI takes over a la Skynet, the huge energy demands for running all the predicted AI use will probably have overwhelmed the grid infrastrucure, overheated the server farms, or broken the internet - always assuming there’s even enough power being created by renewables in our new Net Zero world.
Best take my tin foil hat off now, I suppose; but then as Harold Finch (“Person of Interest” -huh, AI again?) said, only the paranoid survive.
Seriously, though, pob lwc.
I’ve been dreaming about a Welsh course for Russian speakers since I started learning Welsh. Russian is my first language and you can count on me for everything related to it.
Hi Verity - yup, showing words is a slightly sore subject - I would much rather not do it at all, were it not for the fact that I’ve seen it have a very positive effect on the emotional journey for learners who would otherwise be struggling - but I do want us to be proactive about nudging people to try switching it off… things have indeed been hiccupy with AutoMagic, but we’re confident that everything will run more smoothly in the new app - we haven’t prioritised ‘text on/off’ there as yet, but it will definitely get done in the not-too-distant future…
I like your elegant ‘non-request’ for dates 
What I can say is that I’ve just come out of a tech update meeting, and we currently have the end of June as a target for being live in the app stores - that target might slip, but I would be very confident that it won’t slip far.
When the apps are live, we’ll be in a much better position to see what the timelines are looking like for new courses - it’s possible (if everything goes well) that we’ll have 8+ new languages available in-app when they go live - and if that is the case, we will be much better placed to commit to a weekly release schedule, and perfectly happy to put an order of release together
Emma, I’ve flagged that up for Ivan/Kinetic, so you should get a response soon
FrogDoctor (great name!) - it’s all interesting - there isn’t a straight line from ‘let’s use AI to teach languages’ to ‘here’s an AI replacement for SSi’ - and so far everyone who has tried to replicate the SSi method has gone what we would consider a long way off course with some of what we see as the key variables - so we don’t think we’re at ‘copy&paste’ risk of being replaced. It will always be a possibility, of course - but if they do a good enough job and make the world much better at learning languages, we’d have to see that as a win, even if it left us needing to do something else
Anna - diolch! - we will definitely be doing that - I hope it will be before the end of the year, although we’ve got some slight headaches in terms of taking our Welsh content out of the old systems and into the new - it will get done, though, and then Welsh-for-lots-of-other-language-speakers will be one of our biggest goals
Diolch Aran!
Duly twiddled!
This is getting exciting. 
Would it be too optimistic to tell us which 8 those first ones might be?