Former Spotify execs launch AI-powered ‘learning platform’ for the ‘curiously minded’

The oboe establishments know what everyone thinks of artificial intelligence. “Will Amnesty International make us stupid?” The company asks in a recent ad on. “Will we forget how to think for ourselves?”

The founders of the oboe believes that the answer of these two questions noAnd starting them aims to prove this.

The platform of artificial intelligence, which was launched this month, is used by Amnesty International to formulate “courses” on any topic that attracts the user’s approach. It comes from the anchors, and it is the DIY MinI-PODCAST creation platform by Spotify for $ 150 million. Nir Zercharman, CEO of Obi, at a founder, ended in the operation of sound books in Spotify, while Mike Meghano, a Mazouz founder, operated the Podcast team in Spotify.

Zerkman wants to “add a democratic character to reach a great educational experience” cheap with artificial intelligence. freedom. This is a noble claim, especially given the number of products that are described as the already active educational tools-and how the wild hallucinations are still in AI products.

Currently, people resort to ChatGPT, Google, YouTube, Wikipedia and other internet platforms to learn. The process requires users to “assemble it together” and continue to “a very written journey that suits everyone,” he says. Meanwhile, the oboe takes the information to “one destination you need to go to actually learn.”

Oboe has a familiar Chatbot website with a text box that calls on users to write what they want to learn. Unlike ChatGPT, the oboe will not speak with users in a state of astonishment. Instead, the Oboe response to the demand is the “course” created by AI on the subject using traditional educational formats, such as the longest written text that looks like a preliminary semester in a textbook or a short list of “main meals”.

One can generate “courses” about the origin of artificial intelligence or forms of intelligence in nature, both of which appeared in the oboe.

Or, you may be like me and choose the unfamiliar topic of concrete manufacturing and its environmental effects, which I do not know anything about.

I was born this “deep diving” article for me that was classified as a 10 -minute reading. “Imagine a handful of fresh concrete at the palm of your hand,” the section started. The text was dismantled with heads such as “Environmental Fingerprint” and it permeated the tables that included the types of “SCM”. What is SCM? I hit COMMAND + F on the page to define “SCM” but there was no. Pictures of Wikimedia Commons appeared every few paragraphs. After “deep diving”, I managed to click through the information packed in various formats, such as common questions and podcasts.

Although I was away from what I felt was a quick understanding of concrete industrialization, Oboe did not solve the biggest problem in artificial intelligence at the present time: I had no idea whether any of the information was accurate. Confirmation of the details left the learner. The oboe does not include links to the original source materials when saying, for example, that “groups are granular materials such as sand, gravel or broken stone that constitutes the majority of concrete size (usually 60-75 %).” It turned out that this is true according to the American Cement Association, but I had to locate its potential source for myself to verify the statistics.

The oboe is adding this road, but it is not available today. “Certificates and other means to reach additional online resources are things that we are actively working on and we hope to add to the platform in the coming months,” Zerherman wrote in an email comment. freedom.

Other platforms give you quotes when you ask them. When I turned to Google Gemini to get to know the concrete and include it in my requests for categories, I received a link for the 2018 paper in the magazine Nature’s sustainability Accordingly, global concrete production represents approximately 10 percent of industrial water use around the world. useful! Perhaps the categories generally give me a false sense of safety that the information is accurate and is not withdrawn from some AI SLOP. But at least the option is there if you care to search for it.

Zercherman says that Obi does not train his own Amnesty International models, and instead he chooses models from other companies-that he will not call-which are improved in certain tasks, such as mixing speech into the text or large language models. Zekerman says that the mushroom addresses the issue of possible concrete content through the task of some LLMS with the task of correcting other LLMS outputs. In other words, LLMS examines other LLMS.

“For example, you can have one model that comes out of what is believed to be a certain set of facts that must be combined in a cycle, then another model of a different provider who has been trained on a different data collection in a different way that is reviewed, in a way that determines inaccuracy and helps us reduce the possibility of retreat and voices.”

If this fails, there is no feature that allows users to refer to inaccurate content within their outputsand Therefore, at the present time, users need to report inaccuracy directly to the company. Zicherman says they have a “review and integration” process of corrections into the “pipeline”.

If you don’t want to read, Oboe is also born podcast. A couple from the “hosts” generated from artificial intelligence reduced the environmental footprint of the concrete in a similar conversation in NPR’s. Planet money for me.

If the created podcast of the artificial intelligence is similar to the Google notebook, then this is. The difference is that the oboe users do not need it, but they can still download the materials to create a podcast. But this freedom comes at a cost – I trust in the “hosts” of “hosts” from the notebooklm notebooks than a relatively oboe because I know the information comes from my documents – at least, it should be. Meanwhile, the podcast podcasts for Oppo were made of models whose name was not determined, without martyrdom, so I could not bring myself to confidence in what I heard.

The lover will continue to improve as more users demand more “courses”, learning patterns and user conflicts feeding on the models, says Zerherman. It is similar to the oboe to a human teacher, learn how the student learns over time. “Oboe is not just a product that gives you a great dedicated experience, it is also a product that improves whenever you use it more, and the way the human teacher will work while spending more time with them and managed to understand more and more how to learn you effectively.”

Currently, you are still best listening directly from human experts – even if you need to work more seriously to get to know them on the vast ocean of the Internet, increasingly with AI SLOP.