You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
When approaching a new problem, by focusing on easy wins you can sometimes deliver 20% of the results with 1% of the effort. The trap is to extrapolate that the problem will be fully solved if you work on it just 5x longer. Widely applicable to AI successes...
In 1954, IBM managed to do successful machine translation of ~60 Russian sentences into English, using a small set of heuristic rules. Many believed machine translation would be solved in a matter of years. The less you understand about a problem, the more powerful this fallacy
‘If we just get a few more developers, we can solve all the rest of the possible questions anyone could ask our voice assistant...’
In using narrow heuristics to solve narrow problems, we often (and sadly) develop/extrapolate meta-heuristics about how to solve meta-problems.
I bet that the people that worked at the project were not so sure about the “matter of years” part… probably the outsiders could get such an impression, but I can imagine the team to see it failing in a number of trivial cases, grasping the real complexity involved.
But...because there is a “but” here, today nmt has generated a great hype, so expectations are higher than achievements, and if we scale down the problem, well it’s no too far of what happened in 1954...
Not sure why it was ever seriously believed that translation was nothing more than exercise in symbol manipulation.https://twitter.com/fchollet/status/961638043961638912 …
I would like to see how good Google translate is at those same 60 sentences.
In the Goodfellow et al “Deep Learning” book, #MT is included in the ‘solved problems’ part.
Overestimating technology trajectory and willingness of customer to adopthttps://hbr.org/2006/06/eager-sellers-and-stony-buyers-understanding-the-psychology-of-new-product-adoption …
This is one reason why I believe self-driving cars are still decades away from adoption.
the woes of long-tailed distribution. it's amazing in text analytics to be deceived time and again by the fact that rare words convey so much of the nuance.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
