Exploring the new wave of Gen AI Models: Baidu’s Ernie Turbo, Claude and Perplexity | AI in Recruitment Episode 14
In the fourteenth episode, co-hosts Martijn Hemminga and Jasper Spanjaart do a deep-dive into Generative AI models not called ChatGPT: Claude, Perplexity and Baidu's Ernie Turbo. Which ones stand out? AI in Recruitment Benelux Award returns! To kick things off, the episode shared some thrilling news for the Benelux listeners. The entries for the AI in Recruitment Benelux Award are now open for the second time. "This award, inspired by our podcast, recognises innovative organisations in the Benelux that leverage AI to revolutionise recruitment", says Spanjaart. Hemminga added to this excitement by reflecting on last year’s success. "Last year's winner was Knowledgemarker, and we're looking forward to seeing more groundbreaking entries this year. It's a fantastic opportunity to showcase your AI-powered recruitment tools and cases." You can submit your entries until September 5, 2024, via recruitmenttech.nl/awards. Diving into AI Models: Claude, Perplexity, and Ernie Turbo The episode dives into three AI models not called ChatGPT: Claude, Perplexity and Baidu's Ernie Turbo. 1. Claude: Human-like texts Claude's up first. "Claude is an AI language model developed by Anthropic. It comes in three versions: Haiku, Sonnet, and Opus. Haiku is a small, fast model, while Opus is a very large one. Interestingly, the update of the middle model, Sonnet, to version 3.5 made it more powerful than their largest model, Opus, and even surpassed ChatGPT 4.0 according to some specialists." There was praise for Claude’s capabilities, with a nearly human-like text generation. "Claude is quite good at writing easy-to-read texts. The output of ChatGPT can sometimes feel overly influenced by American sales copy, but Claude writes in a way that sounds more human." "Claude can even write simple programs, like creating a quiz about American politics with animations. It's still under development, but it's already very impressive. Claude also has an Android app that might be better than OpenAI’s, alongside its iOS app. That’s intriguing." 2. Perplexity: the contextual AI Model Next, Hemminga introduced Perplexity, an AI model known for its innovative approach to generating human-like text. "Unlike other AI tools that rely heavily on large language models, Perplexity starts by understanding the context of your query," he explained. "This approach sets Perplexity apart." "It emphasises understanding the nuances of your query, conducts thorough research, and generates a response linked back to its original sources." "Perplexity ensures it grasps the essence of your question before diving into the vast sea of information. It emphasises understanding the nuances of your query, conducts thorough research, and generates a response linked back to its original sources. This makes fact-checking easier and more reliable." 3. Ernie Turbo: Baidu’s high-speed AI Lastly, they discussed Baidu’s Ernie Turbo. "This AI model focuses on high-speed and accurate natural language processing. Their website is in Chinese, but with the help of a VPN and Google Translate, we were able to give it some queries." "Maybe in a few years, we’ll conclude that China has won the AI race with Baidu’s various models." "Ernie Turbo is fairly similar to ChatGPT, though it also has a text-to-video feature and boasts 200 million daily queries. It’s the backbone of Baidu Comate, an AI coding assistant, with a code adoption rate of 46%. The images it creates are quite impressive, possibly even better than Microsoft’s Copilot." "Maybe in a few years, we’ll conclude that China has won the AI race with Baidu’s various models", Spanjaart concluded.
