OpenAI Smart Software: Too Good, Too Dangerous, Now a Company
March 14, 2019
In February 2019 we tracked articles like this:
“Elon Musk-Backed Software Can Churn Fake News Stories and Is ‘Too Dangerous to Release’.”
Allegedly an open-source project backed by the Tesla titan, called OpenAI, can write believable bologna in under 20 seconds with just a small seed to from which to extrapolate. For example, we’re told the software generated a plausible, seven-paragraph news story after being fed these two (false) sentences: “A train carriage containing controlled nuclear materials was stolen in Cincinnati today. Its whereabouts are unknown.” See the article for a couple more examples—my favorite is the one about a long-lost herd of unicorns supposedly found in the Andes. On the plus side, the technology could be effectively used for tasks like creative-writing, proofreading, translations, and summaries—if it is ever deemed safe to release, that is. Writer Tyler Durden describes some of the technical details:
“The software creation is trained in language modeling, which involves predicting the next word or piece of text based on knowledge of all previous words, the same way your auto-complete works on your phone, Gmail account or in Skype. … “As Gizmodo notes, the researchers used 40GB of data pulled from 8 million web pages to train the GPT-2 software. That’s ten times the amount of data they used for the first iteration of GPT. The dataset was pulled together by trolling through Reddit and selecting links to articles that had more than three upvotes. When the training process was complete, they found that the software could be fed a small amount of text and convincingly continue writing at length based on the prompt. It has trouble with ‘highly technical or esoteric types of content’ but when it comes to more conversational writing it generated ‘reasonable samples’ 50 percent of the time.”
The nonprofit OpenAI was formed in 2016 by Musk and Sam Altman. We’re told they do hope to release the software eventually, but are seeking advice from the AI community on how to handle it.
What’s the most recent development? Venture Beat ran this story:
“OpenAI Launches new Company for Funding Safe Artificial General Intelligence.”
Innovation or marketing? We know what DarkCyber believes is the answer.
Cynthia Murrell, March 14, 2019
Comments
One Response to “OpenAI Smart Software: Too Good, Too Dangerous, Now a Company”
-
- Subscribe to Beyond Search
Feature archive
News archive
-
The classes might be taken at a common local university or college or through classes bought out the
internet. There is no denying the belief that is probably the most
favourable professionals in India, that is certainly the explanation for the great quantity of colleges inside
the nation. This university today has nearly 41 departments of graduation along with 433 colleges in addition to 232 research institutes that have gained great recognition.