Is Google’s AI Really Sentient? | The Weekly Roundup

Hi there everyone. Welcome to another episode of the weekly roundup your go to for the latest and greatest in data science. This week we're talking about the exceptional funding round of AI startup Papercup and the large amount of discourse that's been started by a Google engineer that claims their large language model Lambda is sentient. So make sure you stay tuned to stay in the loop and don't forget to hit that subscribe button on whatever platform you're watching this Our first story this week takes a look at the exciting new funding announcement from AI dubbing company Papercup. On 9 June, the British company announced that they'd secured $20 million in a series A funding round, making them one of the most funded AI dubbing companies in the world. Papercup is one of a handful of companies using AI to localize content by creating realistic expressive dubs in more languages than content creators could ever manage on their own. Let's take a look at it in action. Hi guys, my name is Margarita and I have a YouTube channel. Hola Chico, Soy Margarita, Y10 Buong Canal and YouTube. The company has said that they'll be using the funding to double down on their research into expressive digital voices and to expand into markets where they've seen great success. The scale at which Papercup aims to dub videos could unlock around 99% of content that's only available in one language, and we can't wait to see where they go next. Our next story is about how researchers have used data science to solve one of the most long standing problems in maximum flow. The problem asks how much material can flow through a network from a source to a destination if the links in the network have capacity limits. So think of something like a highway system or even a job search platform. The algorithm which is explored in the paper linked in the comments has been described as a slam dunk and essentially the best possible solution by Alexander Madri from MIT Satish Rao from UC Berkeley has said that the result made him jump up and down with excitement. Not only does the algorithm devised by the team solve the maximum flow problem insanely quickly, it also allows for a consideration of costs along different routes, making it even more impressive. Daniel Spielman from Yale says, now that we know we can do this really quickly, people will find all sorts of applications for it that they just weren't thinking of before. We can't wait to see what comes next for maximum flow advancements. Our final story this week covers Google's Lambda AI and one of the engineers who built it, Blake Lemoine who's been put on leave after claiming in an article in the Washington Post that the chatbot has achieved a level of sentience. Google's breakthrough Lambda system, which is shorthand for language model for dialogue applications, can engage in a free flowing way about a seemingly endless number of topics in order to better understand the intention of search queries. This is a large improvement on many chatbots used today, which tend to struggle with abstract thoughts and complex questions posed by users. Blake Lemoine published a transcript of the chat he had with the bot, claiming that the AI was thinking and reasoning like a human being. You can find a link to the conversation in the description below. And if you've seen this story popping up on your social feeds like we have, you'll have heard some of the impressive sounding answers generated by the bot. In spite of how human like Lamda might sound, we're still very far away from what could be considered sentient AI. Think of Lambda like a very powerful autocomplete. It returns results that, based on your input, are the most likely to be comprehensible and correct according to the vast amount of text it's been trained on. Imagine someone playing Scrabble in a language that they don't understand, but abiding by the rules to score the most points by stringing together letters that are most likely to generate a word. Gary Marcus, former head of AI at Uber and a critic of the concept that large language models constitute sentient AI, pointed towards a similar story of an NLP system that was produced in 2014, which was also able to showcase sentience by parsing cheering taste. What this story showcases, which is also a point being made by almost the entirety of the AI community, is that we need more strenuous and robust ways of testing for sentience, and this requires a deep understanding of the human brain as well. AI professor Toby Walsh has described Lamda as about as sentient as a traffic light. And Abeber Berhan, a senior fellow in trustworthy AI at Mozilla we have entered a new era of this neural net is conscious, and this time it's going to drain so much energy to refute. Similarly, Erik Brynjolfsson, professor at the Stanford Institute for Human Centered AI, says foundation models are incredibly effective at stringing together statistically plausible chunks of text in response to prompts, but to claim that they're sentient is the modern equivalent of a dog who heard a voice from a gramophone and thought his master was inside. Let us know what you think about Lamda in the comments that's it for another roundup. Thanks for watching. We hope you enjoyed it. And we'll see you same time, same place next week.