Introductionspublished at 13:12
Hope you're enjoying watching Rory multitask
The annual Loebner Prize in Artificial Intelligence is taking place
Judges carry out simultaneous conversations with both a chatbot and a human
They have to try to work out which is a computer program and which a person
An award is given to the most human-like bot entered
If a bot fools half the judges it will win a grand prize of $100,000 and the competition will be discontinued
Leo Kelion, Clare Spencer and Bernadette McCague
Hope you're enjoying watching Rory multitask
Looks like there were some teething troubles.
Rory looks stressed!
But it's started now.
Rory Cellan-Jones
Technology correspondent
I'm just about to sit down. The judges have been discussing our strategies for trying to out the robots.
We've all got different ideas and have all confessed to feeling a bit nervous.
Judges are sitting down and flexing their fingers....
Bruce Wilcox has impressed the judges in the past with his chatbots, which have beaten rivals three times since 2010. He wrote a blog in 2014, external about why his chatbots consistently did well.
His secret, he explained, was that he aimed to create the illusion that his chatbots understood what they were talking about.
“There is no inherent meaning," he wrote.
"Meaning is just an agreement made between the parties to the conversation. One always assumes the other is understanding you perfectly, but the reality between two humans often falls short of that.”
Jane Wakefield
Technology reporter
Yesterday I grabbed a quick interview with Fred Roberts, the creator of a chatbot called Elbot. Elbot won the 2008 Loebner Prize and is, according to his creator "half AI and half a parody of AI".Next month Elbot will launch as an app, offering people an alternative to chatting to Siri, Cortana or Google Now - the smartphone assistants that most of us are familiar with.But Elbot won't be able to schedule meetings or make phone calls, and often he'll be quite rude when he replies.
"He thinks highly of himself and not so much of humans" said Mr Roberts. And don't ask him for help with maths as he is programmed to get the answer wrong. A sarcastic bot - now that is quite human-like!
Meet Mitsuku, Lisa, Izar and Rose. They are the bots that answered questions in the most human-like manner in the semi-finals of the Loebner prize.
Here a selection of the 20 questions they were asked, external:
Later today the judges have five minutes to ask them some more questions to see if they can weed the bots out from the real people.
Jane Wakefield
Technology reporter
Ariadne Tampion is one of this year's judges, but last year she was on the other side - as one of the humans chatting to the judges.
She told me that she tried to answer questions "sympathetically" in that role.
This year, as a judge, she is hoping that she won't be fooled immediately.
"I'd be a bit disappointed if I guessed it was a bot in the first line", she said.
She also plans to "dominate the conversation".
We're keen to hear your comments and will sprinkle some of them through this live page later.You can get involved via: Email:
talkingpoint@bbc.co.uk
Facebook:
https://www.facebook.com/bbctechtent, external
Twitter:
Use the hashtag #intelligentmachines or tweet us at @bbctech, external
SMS:
Send us a text on 61124
No bots allowed.
Jane Wakefield
Technology reporter
Yesterday we came to Bletchley to record our BBC World Service show Tech Tent, and - ahead of the Loebner Prize - I grabbed a quick interview with the founder Dr Hugh Loebner. He told me that one of the most exciting things for him was being at Bletchley Park because Alan Turing is one of his heroes. I asked him what the highlight had been over the last 25 years that the prize had been run. "The first one", he said. He wasn't impressed with 2002 prize in Atlanta, which he described as "chaotic", but he did enjoy the three years that the prize was hosted in his own apartment. "I could roll out of bed, plug in a computer, test the system thoroughly," he said.
Here are more tweeted suggested questions for Rory to ask to see if the answer would give away whether the respondent was human or a chatbot.
Humans may find a few of the questions tricky, to say the least.
Jane Wakefield
Technology reporter
... But the organisers have to be careful to keep the judges separate from the confederates so they have to eat separately.
Jane Wakefield
Technology reporter
I'm not a Loebner Prize judge like Rory, but that doesn't mean I can't talk to a bot.
These days lots of chatbots exist online including a version of one of this year's finalists -Mitsuku, external.
So, I decided to run my own Turing Test and started with a great British conversational opener...
ME: What's the weather like where you are?
MACHINE: Here in Leeds it is quite overcast. I think it will rain later
(Mm - a bit suspicious - not how a human would talk)
ME: Don't put any washing out then
MACHINE: Where should I put it then?
ME: You're a bot aren't you?
We're now streaming live from Bletchley Park.
To watch click on the Live Coverage link at the top of the page.Viewers in the UK will also be able to see the occasional update on the News Channel.
Another deep question has been suggested to Rory:
Meanwhile he has dismissed this question:
Email: talkingpoint@bbc.co.uk
George:
The only true test of a 'thinking machine' is if it uses inappropriate language and tells its creator/tester to go take a running jump.
No truly self-aware machine can be made, because electrons are just a cloud of moving charges.
Our brains work by electro-chemical activity, but if you just replicated the pattern of electrons in my brain in a machine, it would not be me, and it would not be self-aware.
AIML developer and Loebner Prize winner 2013
Rory Cellan-Jones
Technology correspondent
I have just arrived.
The tension is mounting. On the train journey I have been bombarded with all sorts of clever questions to ask the bots. I may use some of them. Watch this space...
Jane Wakefield
Technology reporter
So how does it all work?
There will be four judges attempting to decide whether they are chatting online with a human or a bot.
In a room just behind them will be four humans - known as confederates and four chatbots.
The judges will interact with two entities on a screen and must decide which is human and which is the machine.
There are four rounds and after that the judges must rank the entities as to how human-like they were during the conversations.
Jane Wakefield
Technology reporter
People are starting to arrive now and later on I'll be interviewing some of the judges and also the human 'confederates' who will be busy typing away in a little room behind us.
But first I spoke to Ed Keedwell, a senior lecturer at the University of Exeter who is running this year's competition. He told me that the hope is to host the prize at Bletchley Park permanently from now on - it has previously been held at different locations around the world. Bletchley seems fitting, he told me, because of its links to UK computer science. I also asked him when he expected an AI to win the ultimate $100,000 prize. "I think it is still a long way off - multiple decades", he said.