Danny Miller

I’m using Google Chrome right now, and I have to say that it’s blazing fast. Based on Javascript benchmarks I’ve ran from around the web, I found it to be way faster than IE7, way faster than Firefox 2, much faster than Firefox 3, and even faster than Opera. Chrome’s new Javascript engine is awesome, but I just wanted to clarify that tabs do not always run as a seperate process.

Google mentions on their in one of their Chrome FAQs that it’s up to the web developer to decide if a new link opens as a seperate process or not.

I found this out first hand by going to W3 School’s html samples page and clicking on one of the links. Instead of opening on the same tab, the link opens on a new tab. So now I had http://www.w3schools.com/HTML/tryit.asp?filename=tryhtml_basic open in a tab. I edited the html source code on the left frame and put in a simple infinite loop with Javascript then hit run. As expected, the tab, instead of the entire browser, hung.

However, when I clicked on the W3 School’s html sample tab (which launched the now frozen tab), I noticed it was also frozen. My other tabs, such as Gmail, Google Calendar, etc were fine.

Google answered why: 

New tabs spawned from a web page, however, are usually opened in the same process, so that the original page can access the new tab using JavaScript.

Makes sense. In fact, that is one of the reasons why current web browsers run in a single thread. No big deal. Most of the time I have multiple “sessions” of browsing and I don’t mind if a few tabs share the same process thread.

Anyway, if you don’t have Google Chrome, get it. Finally multi-core machines will see benefits in browsing the web. I can have 10 tabs open without seeing a slowdown.

Today Google reported that they’ve developed an algorithm to index Flash content (only text, not video or images). This news is clearly dominating Flash news sources around the web, with mostly mixed reviews.

Many bloggers are criticizing Google claiming that it’s impossible for any algorithm to figure out the text information of SWF files which load text from external sources (such as XML) because it’s impossible to know the format of the XML documents being transferred over. But who cares? I still don’t get why people are saying that that’s how the algorithm works. Google has stated that it’s able to crawl externally loaded SWFs (although they don’t couple it with the original SWF when indexing, which is a significant problem for sites that load multiple SWFs for navigation); consequently, they must be monitoring HTTP requests made by the SWF and can do the same with XML files. Google doesn’t need to know how the XML file is parsed… the Flash document will do that for them. They can just have the Flash load the XML file and monitor the text fields and see the value. That’s probably why they say, “To protect your material from being crawled, convert your text to images when possible”.

The only problem I see is if text fields are very dynamic. Maybe the algorithm only goes through static text fields? Because I see no way how a text field that displays random letters (for visual effect) being able to be indexed by any algorithm.

Here’s my prediction: Community tagging. Just like the Google Image Labler game, Google will ask users to tag/label Flash documents that their parser can’t index correctly. Humans would be the perfect computational tool to solve this kind of problem. Yes, there are millions of SWFs out there needing to be indexed, but we really just want the major ones parsed.

The keyboard bug with RC1 has been fixed for latest build (RC2)!

Phew!

The Mozilla team has announced that they are attempting to set a (Guinness Book) of World Record of number of downloads for a piece of software in 24 hours… A record never set before.

The first thought that came to me was… huh? How can this number not be set already - by Adobe (or back in the day Macromedia) Flash! Isn’t the Flash plugin downloaded millions of times per day? Flash is on more computers than any other software (including any operating system). Macromedia never and Adobe doesn’t really release any download statistics other than averages. I’m sure Flash 8 or 9 when they first went public shattered records. Adobe should release the results so Mozilla has a number to beat.

Just think back to the day YouTube decided to switch to Flash 9 and how many users downloaded then.

Mozilla may set the record, but by no means will they have broken one.

*Edit: Mozilla targets 5 million downloads within 24 hours, but I guess the didn’t know that Flash Player 9 is being installed at the rate of about 12 million a day.

I love FF3 (although I am shocked that they won’t fix a critical Flash bug that happens with RC1 for the final release). The search bar is great, and I don’t experience the slowdowns I did with FF2. But still, there seems to be a disconnect with FF and the web… Just look at this huge bugzilla thread about the Flash bug.

Here are some highlighted features of the new player (a list by Adobe can be found on their labs page)

• Adobe has finally made noise (read introduction material parts 1, 2, 3 by Adobe Engineer Tinic Uro). Keith Peters (who is one of the lucky few to have a version of Adobe’s upcoming authoring tool) has posted a sample application showing dynamic sound.
• Native 3D effects! Interesting to see how Away3D and Papervision will react. While the native 3D addition will be great for vector graphics, Astro doesn’t support texture mapping or 3D model importing from 3rd party software while current open source projects like Away3D and Papervision do.
• Multi-column layouts/tables for textfields.
• Ability to change bitrates for streaming video on the fly.

Here are some demos by Adobe.

My reaction: After a survey showed some astronomical 98% of online videos use the Flash Player, Adobe seems like they are trying to cater to those needs. There are HUGE additions to video with Astro, and while I don’t see video portals such as YouTube or Google Video using the native 3D engine (unless it’s for some visualization), I do see web designers smiling as they can now deliever unique features for their clients. It’s only a matter of time before a 3D navigation using video is released by some design firm for a client.

There are also some… somewhat random… additions to the Flash Player. For example, inverse kinematic support with a new “Bones” tool. Is this feature really necessary? What was wrong with just using one of the many 2D physics engines?

Unfortunately, Adobe hasn’t released LiveDocs yet… But that will come soon. Understandable since much of the syntax is subject to change.

I wonder how long now before AS4 and the new ECMA features…

This year, only three entries were submitted in the annual Loebner Prize Competition. The prestigious $100,000 dollar prize has yet to be won, and the competition is the only place where it has a chance to be awarded. If a team develops a chatterbot that can fool judges into thinking they are communicating with a human, that team will win the prize and the prestige. So why were only three entries submitted to this year’s 17-year-old event? Have developers given up? The answer is yes and no. While developers are still working to expand the intelligence of chatterbots, the focus is no longer on fooling humans. With the rapid advent in information technology, chatterbot developers are adjusting focus to keep up. The 1966 chatterbot ELIZA spawned this specific field of human-computer interaction; however, today’s chatterbots no longer have the same goals as ELIZA. For example, one of the latest chatterbots named SmarterChild has branched from ELIZA from the field of artificial intelligence to information technology. While ELIZA and the chatterbot SmarterChild are similar in that they mimic human behavior and response, they greatly differ in purpose and behavior.

ELIZA attempts to act as a Rogerian psychiatrist by continuously asking questions to its users. A user types in text as an input into the ELIZA program, and ELIZA returns a response that, ideally, is coherent and sensible to the user. The first testers of ELIZA sent messages via teletype to ELIZA not knowing that they were chatting with a robot instead of an actual psychiatrist. Shockingly enough, ELIZA was able to fool a great deal of its users into thinking they were communicating with an actual human psychiatrist. In fact, ELIZA was able to elicit deep, sensitive responses from its users. This artificial intelligence field has since lead to new, more sophisticated chatterbots with the same goal of fooling and gathering sincere human response.

Alan Turing had some assumptions about mankind. By now, he expected chatterbots to be able to fool most humans into thinking they were chatting with another human after a few minutes of conversation. Unfortunately, in terms of the Turing Test, we are not even close. Even the latest award winners of the Loebner competitions yield bizarre responses after one or two sentences. Does this mean chatterbots are lagging behind Turing’s vision?

Maybe bots have just taken a step sideways toward a different vision altogether. SmarterChild, a chatterbot developed within the last few years, is arguably one of the most sophisticated chatterbots available. Like ELIZA, SmarterChild can mimic human response. However, SmarterChild is highly developed in its responses in comparison to ELIZA. For example, a typical conversation opening with ELIZA may produce the following response.

You: How are you?
Eliza: Does that question interest you?

Compared to the more refined memory of SmarterChild:

You: How are you
SmarterChild: I’m doing great, how are you?
You: Ok
SmarterChild: OK? Glad to hear it!

Nevertheless, the reason SmarterChild is one of the most advanced chatterbots is not because it can impersonate human dialog extremely well. SmarterChild is advanced because it has a huge database of information resources, it remembers individual human response and bases it in future responses, and changes personality based on “peer pressure” with its audience. SmarterChild is able to return movie schedules, weather forecasts, news, opinion polls, random statistics, trivia, and tons of other types of information. The first pieces of information SmarterChild asks for from strangers are name, age range, and zip code. SmarterChild uses this data for local news information, weather, news, and more. Additionally, SmarterChild changes its own views and questions based on opinions inputted by its users. For example, if users tend to like Pizza, SmarterChild will change its favorite food.

In terms of credibility, SmarterChild does not offer an experience like ELIZA. Credibility implies believable emulation of a particular persona. The function of SmarterChild is to be a friendly encyclopedia rather than a credible, stereotypical personality. On the other hand, ELIZA is indeed a character. ELIZA is programmed to be someone and respond in a certain way to fulfill a specific character or stereotype to produce a believable experience. No matter what the user inputs, ELIZA will pose a question in the form of a Rogerian psychiatrist.

ELIZA creator Joseph Weizenbaum argues that no robot can ever replace humans because humans have the ability to comprehend or exhibit emotion. Yet, the implications of ELIZA have drifted away from emotion and feeling comprehension. Yes, users of Microsoft Office may possess negative feelings of hate toward the Paper Clip “help” agent; however, the Paper Clip is unable to respond to those emotions and can only react based on internal states and decisions made by the user. For example, the Paper Clip can be dragged around and silenced, but it will not sense that the user is upset when it repeatedly asks whether the user needs help writing a letter. ELIZA was built to challenge this idea of comprehending emotion, but it does it in a very superficial way.

You: I am sad.
Eliza: Do you think coming here will help you not to be sad?

Unlike ELIZA, SmarterChild is more of a tool than a character. Like a search engine, email client, or word processor, SmarterChild helps the user accomplish tasks. Yet, SmarterChild has a different approach to information technology than other Web 2.0 projects that attempt to do the same tasks. Contrast to websites, there are no “menus” to access information in a hypertext only environment. SmarterChild is a chatterbot. Because of this, SmarterChild can deliver information to its users faster than any website. The user types in the desired knowledge in the chat window as a sentence, and SmarterChild relies on its linguistic artificial intelligence to understand the context and what kind of response to deliver. Additionally, because SmarterChild is on an instant messaging medium, it can write hyperlinks to allow users to quickly access websites.

You: What movies are playing?
SmarterChild: Movies playing in or near Marietta, GA (30006) on Tuesday, October 23rd:

1 The Game Plan [PG]
2 30 Days of Night [R]
3 The Comebacks [PG13]
4 The Heartbreak Kid [R]
5 Michael Clayton [R]
6 We Own the Night [R]

Type more for the next set of results.

ELIZA has nothing to do with information technology; it was built to progress artificial intelligence and Alan Turing’s prophecy. But today, the goal is not to build a robot to fool a human. One key difference to SmarterChild and ELIZA is that SmarterChild is aware of its nonexistence.

You: Are you a robot?
Eliza: Would you prefer if I were not a robot?

Versus SmarterChild:

You: Are you a chatterbot?
SmarterChild: Yes indeed.
ScratchFromStart: Are you human?
SmarterChild: No I am not.

In this day and age of knowledge, the goal is to make a robot to deliver information in a relative, contextual manner.

The revolutionary affect of ELIZA was not the subsequent emergence of human computer interaction and innovations in chatterbots. The real influence came in information technology and discovering new ways to deliver information and knowledge to users in a friendly, humanistic style.