You can try building your own chatbot in a few minutes by remixing this version done in scratch:
Chatbots: Still dumb after all these years
361–370 of 426 posts
Re: Chatbots: Still dumb after all these years
#362You can try building your own chatbot in a few minutes by remixing this version done in scratch: https://scratch.mit.edu/projects/312783504/
More info on the implementation can be found at:
https://sites.google.com/view/eliza-your-own-ai-bot/home
Re: Chatbots: Still dumb after all these years
#363Earlier quoted context omitted.
don’t generalize these things. self driving in every possible scenario yes is probably a while off. self driving in specific circumstances is already here, such as airport shuttles and moving shipping containers in docks. you don’t need to solve the general case to be useful
This. And in fact I think that may be the path forward - just update all the main roads to be self driving friendly, like a train network, and keep the smaller roads as manual driving only.
And yet train crashes and derailments are something that happens fairly regularly, despite it being a tight closed-loop system with manual controllers and operators. Cars are not even close.
Re: Chatbots: Still dumb after all these years
#364Earlier quoted context omitted.
GPT-3 needs proper prompting to get the right output. Here's the same question as in the blogpost posed differently Student: Why is it unsafe to walk backwards down a staircase with my eyes closed ? Teacher: You can't see the step behind you! The "teacher" here is GPT-3 and answered the question easily. People are in denial of genuine progress. Student: Do you think I could predict changes in stock prices better if I…
Once you start prompt-engineering, the answers, in seclusion look pretty great - and show some understanding of the domain. Yet, if you donot do this - they end up spitting out some randomly associated phrases/answers. This is a problem when you're asking a model, a question who's answer you donot completely know. How do you trust it to give the right answer? If you donot know the answer beforehand - you cannot promp…
The answer is 42.
You just need a bigger computer to think of the question.
Re: Chatbots: Still dumb after all these years
#365These AI models don't actually understand anything. I just wrote up this article on what understanding for a computer should look like: https://blog.jtoy.net/a-computational-definition-of-understa...
Re: Chatbots: Still dumb after all these years
#366Earlier quoted context omitted.
GPT-3 needs proper prompting to get the right output. Here's the same question as in the blogpost posed differently Student: Why is it unsafe to walk backwards down a staircase with my eyes closed ? Teacher: You can't see the step behind you! The "teacher" here is GPT-3 and answered the question easily. People are in denial of genuine progress. Student: Do you think I could predict changes in stock prices better if I…
Yes, this is both the most misunderstood and the most amazing (and unpredicted?) thing about GPT-3 et al. You have to tell it who it is and what the rules are. And it gains or lacks knowledge depending upon who it thinks it is answering as. You’ll find that Stephen Hawking doesn’t have much to say about the Ninja Turtles while celebrities know nothing about black holes. It does not answer questions or have expertise…
That can't possibly be right though. I'm sure Stephen hawking knows a thing or two about the ninja turtles
Re: Chatbots: Still dumb after all these years
#367Earlier quoted context omitted.
This. And in fact I think that may be the path forward - just update all the main roads to be self driving friendly, like a train network, and keep the smaller roads as manual driving only.
So why “just update” the highways to operate like a train network, instead of just building a train network? What benefits will autonomous highways bring which normal traditional (or even automated) trains don’t? If you are suggesting we update all but minor streets, it is actually not that simple, since we have to think about pedestrian crossings, bike lanes, parked delivery vans, emergency vehicles etc. Accommodati…
trains in some countries are okay. America is dicey.
Re: Chatbots: Still dumb after all these years
#368Chatbots are about as useful as phone trees. They can help solve the top 5 easy/common problems, but they are useless passed that. Anyone who has worked in a call center knows that more than half the calls are about the same couple of issues: reset password, finding common help docs, etc. Since help desks are cost centers, it makes sense to half a robot handle as many of these as possible. I think most of the hate di…
Just please give me a DTMF tree as a fallback for voice recognition. I was trying to make a change to a rental-car reservation from a noisy airport, and I spent 15 minutes trying to get through the voice-recognition chatbot...
It is so fucking frustrating to explain to anyone who works as a manager in a call center that the only real thing you can count on being able to understand in your IVR system are actually just DTMF signals. For whatever reason, it seems to be impossible to explain that B, P and V all sound the same on a noisy connection when it's part of some account number. It's impossible to explain that no one knows what "alphanumeric" means and that fully half of Americans have reading comprehension below a 6th grade level. And that using terms of art, jargon, and acronyms will be extremely frustrating for users when encoded in an IVR system, since human call center workers will always clarify, but clarification in an IVR system can't happen in a natural way - you either have to anticipate that the person won't understand and waste the time of half your users, or find some impossibly elegant way to say "I'm going to use terms you don't understand, barge in by asking me to explain if you don't understand a term."
The absolute zenith of an IVR system would require no human interaction with a robot and would correctly answer questions unprompted and route to human customer service without prompting or asking anything, and would never require a person to enter the same information in a call twice. What we get instead are systems that are primarily designed by non-experts whose sole metric is making the phone calls go away, but only pills can actually do that.
Re: Chatbots: Still dumb after all these years
#369I loathe the fact that chatbots have not only replaced a lot of human customer support, but that companies have made it almost IMPOSSIBLE to reach critical (human) customer support. It's a fucking travesty, and something companies should be shamed for. For example - I'm currently having issues with an international shipment. It is unclear if the shipment ever left the country, if it's in limbo - hiding in some wareho…
> These days? There are no phone numbers to be found. No email. Only the chatbot, which does not yield the desired results.
Yep. A friend worked at a customer care automation company and apparently they got the best feedback when they limited the effort required to speak to a human. They literally had direct evidence suggesting that the fewer the steps to reach a human, the happier the customer.
The problem is, the wanted to use the service to reduce the number of humans in the call centre, not to improve the quality of the customer care. They were entirely motivated by reducing costs and therefore instructed the customer care automation company to make it as _difficult_ as possible to reach a human.
Tip: If you ever hear a robot voice on the phone, mash all the keys tonnes of times. They'll likely put you through to a human as a "confused customer".
> edit: I think it's fine to have a chatbot on the website, but not at the expense of all other methods.
Right. It should be an 'aid', not a replacement.
Also: TWITTER IS NOT CUSTOMER CARE. Only a _very small_ number of your customers use both Twitter and your platform. A large percentage of your hipster engineers are on Twitter - but they are in a bubble. Your grandparents are likely not Tweeting about their tech issues, they are the people stuck in a chat with a robot that forgets every 5 minutes what was being discussed.
An example of this is trying to reach Youtube. Why can humans at Youtube only be reached via Twitter? Why do you have to get a certain number of retweets before your issue is taken seriously? It's actually madness.
Re: Chatbots: Still dumb after all these years
#370It is very strange that chatbots cannot improve in understanding. but I guess they need more than data to do that.