Learning a Language in the Future? I asked Claude+

Open Forum
1

I’ve asked Claude+ to give me some strategies for the future of AI and learning languages.
Then I asked to give me some predictions about time and money saved, and if humans will study more languages.

I think it gives an interesting answer.

I personally believe that more people will have easy access to learn a new language, there will be less barriers, and it’ll cost a lot less money. What do you think?

POTENTIAL STRATEGIES
Here are some potential foreign language learning strategies using advanced AI in the future:

• Personalized interactive lessons with an AI tutor. A customized virtual tutor or coach could guide a student through highly tailored lessons, conversations, and practice activities based on their needs and progress. The AI could get to know the student’s strengths, weaknesses, interests, and motivations to provide a personalized curriculum.

• Immersive augmented/virtual reality environments. Students could travel to realistic simulated environments where they can physically experience scenarios requiring communication in the target language, with AI characters for conversing and practicing with.

• Continuous comprehension and speaking practice with AI agents. Using natural language processing, students could have fluid extended conversations, discussions, and debates on various topics with AI entities designed to build speaking and listening proficiency through meaningful practice.

• Instant feedback on speaking with advanced speech recognition. Systems could provide detailed, nuanced feedback on pronunciation, accent, fluency, vocabulary, grammar, and more during conversations with AI agents or in immersive environments. feedback could also estimate overall speaking ability and fluency.

• Automated error correction and writing feedback. AI could review students’ writing, such as essays, blog posts, or messages, and provide specific feedback on vocabulary, grammar, syntax, punctuation, and colloquial language to improve proficiency.

• AI-generatedBooks, video, audio for language learning. Customized books, short stories, scripts, video lessons, podcast episodes, or other multimedia content could be generated on student’s areas of interest at the right level for learning. Interacting with this content helps language exposure.

• Personalized content curation based on interests. An AI system could curate or recommend authentic content like news articles, books, movies, TV shows, music, podcasts, social media, and more based on a student’s interests and hobbies. This helps supplement learning with enjoyment.

• Gamified challenges and quests. Interactive games, challenges, missions, and quests developed by AI could provide students with motivation and engagement using storylines, characters, and experiences personalized to their needs. These help to learn while having fun.

• Social connection with others. AI could help match students with potential language exchange partners or tutors around the world based on interests, motivations, personalities, and skill levels to facilitate rich collaborative learning and culture sharing.

PREDICTIONS
Here are some predictions on how AI could impact language learning in the future:

• Time saved: A person could save up to 50-70% of time studying a language using advanced AI for personalized, optimized lessons; instant feedback; conversational practice; and other automation.

• Money saved: Someone could save roughly $500-$1000 or more per year on average using AI instead of traditional courses or tutors. The more AI is used, the greater the potential cost savings. Probability of at least 50% savings is high.

• Increased access: Due to free or low-cost AI apps, at least 50% more people worldwide will have the ability and resources to learn a new language. Probability is very high at 90% or more.

• More languages learned: There is a 60-80% chance that the average person will learn 2-3 languages in their lifetime instead of 1-2 languages today due to the ease of learning with AI. However, there is also a 20-40% probability of more standardization around the most popular languages.

• Will people learn more languages? Likely yes, with a 70% probability. AI will make learning more accessible, personalized, and efficient. However, some may focus on just 1-2 languages if not needed for work or travel.

• Standardization probability: Around 30-40% chance of more standardization as certain languages like English, Spanish or Mandarin become more prominently used in business, science, and culture on a global basis. But niche or regional languages will still remain important to many.

• Time saved per level: A student could save ~25 hours (50%) to reach an intermediate level and ~50-100 hours (60-70%) to reach an advanced level using AI versus traditional learning methods. Probability of at least 50% time savings at all levels is very high at 80-90%.

• Money saved per level: Estimated $200-$500 (50%) savings for intermediate and $500-$1,500 (60-70%) savings for advanced using AI. Probability of at least 50% savings at all levels is high at 70-80%.

• Increased access per level: AI could provide intermediate learning access to 50-100% more people and advanced to 100-200% more people or higher. Probability is very high at 80-95% for significant increases at all levels.

6 Likes
2

Didn’t know about Claude+. I’m already getting kind of dizzy trying to keep track of all of the AI tools and extensions for them coming out everyday over the last 6 months. I need a separate AI assistant who will do that for me.

3

ChatGPT has already saved me from spending 500 euros to learn Grammar that is covered at the B2 level(grammar concepts like nominalisierung und verbalisierung) here in language schools, Germany. Its explanations were really good. Private language schools charge 500 euros per level and lasts for 3 months. AI+LingQ is a great combination.

4 Likes
4

I actually stopped taking track of those because as you realized there are too many and it’s a real waste of time, unless you don’t make out of that tracking. At this point, if I need something, I might ask on some Discord chat. There are too many changes too fast! :open_mouth:

1 Like
5

Well, in this case, if you have that money saved, you could pay for gtp-4 plus. It’s worth it if you use it often.

1 Like
6

How’s it going in Italy with the recent governmental restrictions on ChatGPT? Do you have to use VPN for the access to it?

7

It’s still not accessible from the main site but it’s working on any other platform that is using their API and it’s still working on Playground. Hopefully it’ll be just something that will be fixed in a few days or so!

2 Likes
8

my only gripe with old version it cuts off the story sometimes without concluding it.
I use it a lot to create stories with newly encountered vocabulary that I come across while reading. I also create fill in the blank type exercises and example sentences based on particular grammar points. For my purpose, it does its job efficiently. if there is an improvement then I will consider it

1 Like
9

gpt-3 is capped at 4000 token (included your request and the answer).
gpt-4 is 8000 token

But you can ask to continue the story.

gtp-4 is a huge improvement.
I mean:
gpt-3: 175B parameters - 40GB of text
gpt-4: 5.5 trillions parameters - 45TB of text

2 Likes
10

What about Google’s Bard? How does it fare against ChatGPT-4? The Bard will be incorporated with Google search engine as well.

11

I don’t know yet, it’s still not public like chatGPT. Having it incorporated inside a search engine is not the same that having it like we are using these tools. I believe we’ll see it soon. There are still even way more powerful in the background but they are like a wild horse. :smiley:

12

I’m on the beta (if that is what it is) for Bard. I tried some of the mnemonic things today. It did ok with the give me a “mnemonic”, etc. etc. Actually produced the same one as chatgpt.

I tried david’s more sophisticated mnemonic query for examples of all the meanings and give a story. It failed this pretty much and just gave the meanings and then the story, but no mnemonics.

I’ve not tried anything else yet with it.

2 Likes
13

Hm, if you know “what to do”, you can learn most L2s basically for free right now (!) just by using the Internet, i.e., Memrise, Anki, YT vids / free podcasts, deepl, ReadLang, Wikipedia, news sites, TV stations, writing forums, etc.

You don’t even need LingQ & Co for SLA - it’s just more comfi.

“Will people learn more languages?
Likely yes, with a 70% probability.”

No - why would they? Language learning is extremely time-intense, esp. when it comes to distant L2s.

Most of our co-human bros and sis will probably skip the learning - and let the AI (translation tools) do the heavy SLA lifting (even if it’s imperfect).

One thing AIs still have to learn is this: “One should never - and I mean: NEVER - underestimate the human propensity for physical and / or psychological convenience.” :slight_smile:

2 Likes
14

One thing AIs still have to learn is this: “One should never - and I mean: NEVER - underestimate the human propensity for physical and / or psychological convenience.”

ha ha ha. You ruined all my enthusiasm. But I’m more confident on the future and on that note: I go to my comfortable bed now! And I’ll have an improbable good night of sleep! :slight_smile:

2 Likes
15

I think some sort of AI language partner that perfectly knows your level in a language and can auto generate short stories to put every word you’re struggling with in an interesting story for you would be an amazing program

4 Likes
16

You could probably do it already right now as a workaround. Go to LingQ, take a bunch of words with the level you want to, ask to create a story with those words, or a mnemonic like Eric did in the other thread. Then you could go to an image program and create very easily a bunch of images to make your “1 page comic story”.

2 Likes
17

“You ruined all my enthusiasm.”
Well, I’ve had a few hundred learners in my life (SLA, math and programming). Many of them (including adults) could only memorize things or imitate others (teachers, siblings, parents, friends), etc., but had given up thinking for themselves. I called this “learned helplessness.”

For example:
More than 80-90% of my older students (16-20 years old) couldn’t solve a simple calculation like 1/2 * 2. The result was always: 2 or 4 or the plea “Please, don’t be such a sadist and give me a calculator. This is way too complicated for me” (in plain English: it’s way too complicated to cut a cake in half and then put the halves back together to have a whole cake - for someone who wants to go to university/Fachhochschule and make a lot of money (I’m not sure as what: a naked influencer?).

If you talked to them for a few hours, you knew they weren’t stupid, but they were so addicted to their calculators that they had completely given up on thinking for themselves.

AI will make this tendency much, much worse - at least that’s my prediction.

4 Likes
18

– AI will make this tendency much, much worse - at least that’s my prediction.

If AI development keeps up the current pace, we all will soon become naked influencers at best. Not a bad idea to at least get one’s butts in shape until then! :smiley:

19

We’re probably going to know which of these (if any) is correct sometime in the next six-months to a year. I notice that the article didn’t mention personalized content targeted to the specific vocabulary you want to learn.
I’m personally hoping that’s the secret sauce to get me out of the grind.

20

With the public APIs anyone can build its own version of AI with the data they have. These are big generic bots but you can narrow them down to your needs. I don’t care about doing it but you can potentially doing it already now, and create your own personal chatgpt with free software on github. I mean, for yourself in your own computer! Playground just went Open Source if I’m not wrong.

Potentially, you can limit it only to the words you want to or exclude the ones you don’t want to from the source.

2 Likes