“He has an edgy humor,” said Iran Reyes, vp, global head of engineering at Media.Monks. “He has a specific communication style. Its answers are short [and] there are some personality traits…He knows how to answer things. It’s not scripted. It just gets an idea of how to answer things and and based on that he’s able to [answer on its own].”

Wormhole is a physical manifestation to showcase Media.Monks’ new AI platform called Monks.Flow, which debuted earlier this month during CES. The platform is designed to help integrate a number of AI tools across various large language models and data sources to deploy chatbots, automate processes and use generative AI capabilities in a number of ways.

Created with the help of a Media.Monks copywriter, Wormhole is essentially an AI advisor as expressed by a sassy alien and created using a number of AI platforms. Dialogue is powered OpenAI’s speech recognition model Whisper, and the voice was created with Amazon Polly, a text-to-speech platform. And rather than relying only on ChatGPT, Media.Monks created an internal tool for switching between various LLMs. For example, Wormhole ran on GPT-4 for demos at CES, but used used Meta’s LLaMA 2 and Amazon Bedrock when it debuted in late November at AWS:Reinvent.

Media.Monks is also able to change the personality and provide various prompts and additional data, depending on the situation. The bots made with Monks.Flow can also be connected to various sources of information. For example, Wormhole was connected to search Google during CES but can also pull various PDFs or Google Drive depending on the knowledge it is searching for.

Wormhole isn’t meant to just exist in the real world; its digital siblings can also be used as AI agents for other tasks. And since it’s not directly connecting Wormhole or other bots to an LLM, Media.Monks can use various guardrails to make sure it’s fetching information from the right sources or saying anything that’s inappropriate. That control makes the bots safer – at least in theory — and also more customizable.

“That was the main goal when we created it,” said Rafael Fittipaldi, partner & global svp of innovation at Media.Monks. “Agents work for brands as well. So it could be your sparring partner on a creative session. It could be a bot to generate e-commerce solutions automatically where it knows everything about the product and people can ask about it.”

When Digiday interviewed Wormhole at CES, the alien often took a few moments to ponder a reply as a screen next to it showed how it was searching the internet in real time based on the question while quickly querying results based on a range of related keywords. Some answers were fairly generic, while at other times it didn’t quite seem to understand questions that would have been easily answered by ChatGPT. (Some responses might have have flawed because of Wifi issues inside the Las Vegas hotel suite where the demo took place.) But overall, Wormhole had plenty of personality and at times was still very spot-on.

When asked about trends at CES trends, Wormhole said to “Brace for a barrage of buzzwords: metaverse, digital health, food tech, blockchain/web3, robotics and AI, self-driving vehicles, VR and AR, and don’t forget new shiny gadgets.” But when asked about journalist Marty Swant, it described him as a “Digiday darling and AI-chatting hot shot always burning the midnight oil in Brooklyn and asking the toughest questions. Quite the character.”

Other AI news:

• The snowy alps didn’t cool off the generative AI discussions last week during the 2024 World Economic Forum in Davos, Switzerland. Last week, world leaders, tech execs, marketers and AI experts discussed a range of AI topics including how to safety build AI tools, and how to mitigate AI-generated misinformation around the world. WEF’s AI Governance Alliance also released a new report and several papers in collaboration with Accenture and IBM Consulting focused on global collaboration for building safe AI. 

• OpenAI previewed it strategy for preventing its AI models, and platforms like ChatGPT and DALL-E, from being used for creating and spreading misinformation ahead of the 2024 election. The company, which mentioned plans to use news and image verification programs, also said it will have “more to share in the coming months.”

• The head of the U.S. Securities and Exchange Commission warned companies to avoid making false claims about their AI’s capabilities. Last week, SEC chairman Gary Gensler also said companies should make sure to address the risks of AI and warned that using AI to mislead the public is still against the law. Gensler’s comments, reported by National Law Journal, came during a speech at an event hosted by Public Citizen, a nonprofit think tank.

Prompted Products: Other AI-related announcements 

• Samsung announced its new Galaxy S24 smartphone and a new Galaxy AI platform to power new AI tools for mobile devices. It also announced partnerships with tech giants like Google and Qualcomm, with the former powering new Galaxy S24 features with its Gemini model and the latter powering AI with its latest Snapdragon chip.
• Google also debuted its new Circle To Search feature, which will let Android users search for anything on their phone without switching apps.

• Meta CEO Mark Zuckberberg revealed that the company is investing heavily in AI with plans to spend billions of dollars on massive amounts of computing power from Nvidia. It also plans to integrate its AI research team into the company’s product team, according The Verge, which obtained an internal memo about the changes. Zuckerberg is also reportedly planning to create open-source artificial general intelligence (AGI) — something that many researchers say isn’t yet possible.

• Researchers at the University of Chicago released a new tool that artists can use as a sort of poison pill for AI image models. The tool, Nightshade v1.0, was originally announced last fall in the form of a research paper but made available for use last week.

• A number of companies released new AI products for marketers and creators. GroupM’s Acceleration added new “AI Maturity Model” for marketing, the AI startup Glystn announced a new AI product for creators. Meanwhile, Rabbit, new AI device startup, announced it is working with Perplexity’s large language model to power the new Rabbit R1 device, which made a splash last week when it debuted at CES.

• CUNY announced a new $75 million donation from the Simons Foundation — the largest donation in the university’s history — with $50 million going to fund computational science and another $25 million devoted to AI research. The funding follows New York state governor Kathy Hochul’s early January announcement to create a $400 million public-private investment for AI. Other higher education institutions that will received funding for AI include New York University, Columbia University, Cornell and the State University of New York. 

Other news from across Digiday