ChatGPT, Lensa and the AI ​​era for everyone

No longer sublime models, AI tools such as ChatGPT, Lensa are increasingly interesting and simple, anyone can access and use.

ChatGPT is the AI ​​that has been mentioned the most in recent weeks because of its ability to answer questions, write texts, build scripts ... according to users' preferences. Users may ask "Who is George Washington Carver?" and get answers from Wikipedia similar to Google. However, its power is greater than that, allowing unlimited creativity. For example, if the request "write a screenplay about a taco sandwich fighting a hotdog on the beach", there will be no answer on Google. But ChatGPT immediately returns results in seconds.

Illustration for AI ChatGPT. Photo: Erick Butler/Unsplash

The WSJ tried putting this question in ChatGPT and got dialogue that was rated as funny as a YouTube video, if not in the style of Netflix movies:

- Taco: "You think you can beat me Hotdog? You're nothing but a flavorless meat product."

- Hotdog: "Okay, you can boast of being made with fresh ingredients, but I have the advantage of being able to eat with one hand."

Unlike Google, the above result will have different content if the user clicks on ChatGPT's reply button again. In fact, ChatGPT does not search, but the response is analyzed, filtered and personalized based on the huge data collected from the Internet, then trained for this tool.

Before that, AI tools like Dall-E or recently Lensa also made a splash when serving user requests with just one click. With Lensa, photos will often become more artistic, beautiful, and sometimes a bit funny.

These software products represent more than just advanced AI tools. More importantly, they make AI from something sublime become more accessible, applied in a simple way to life, especially for non-computer experts. "Six months from now, you'll see amazing things you've never seen before," said Oren Etzioni, CEO of the Allen Institute for AI, a nonprofit dedicated to AI research and engineering.

ChatGPT, Lensa and AI for Everyone

ChatGPT, announced by OpenAI on December 1, evolved from the GPT-3 prototype but is trained to give feedback like a conversation. Originally GPT-3 only predicted passages that would appear after a certain sequence of words, but ChatGPT evolved as it sought to interact with user questions in a person-to-person fashion. The result is answers with the same fluency as real people and the ability to converse on a wide range of topics, representing a huge improvement over the chatbots that appeared a few years ago.

Unlike regular chatbots that follow a pattern, ChatGPT can reply or decline to reply based on context, as well as give new, idiosyncratic responses. For example, in response to the question: "Who am I?", ChatGPT said: "I cannot answer who you are. Only you know and define yourself".

"It's almost like a tool to help people brainstorm and tell themselves to think differently," said Sarah Hoffman, vice president of AI and machine learning research at Fidelity Investments. He tried using ChatGPT to compose a sample research paper and included some passages by other people. In response, the tool warns: "The article may be from 5 years ago".

For programmers, ChatGPT also helps by offering difficult coding solutions. This is the experience that Javi Ramirez, a 29-year-old software developer in Portugal, has gone through. "It saved me," Ramirez said. "An hour of Google search solved in just 5 minutes on ChatGPT".

Meanwhile, from November, the Lensa AI art tool also caused a fever on social networks with the Magic Avatars feature. User-uploaded photos are modeled in various art styles. The app from Prisma Labs uses the Stability AI model to train and produce works that delight users in just a few taps. Users only need to upload 10-20 photos of themselves, the algorithm will process the images to produce portraits in a variety of styles such as sci-fi, fantasy and cartoon. The app offers a 7-day free trial, then charges $39.99 per year.

Threats from the new generation of AI

According to experts, tools like ChatGPT or Lensa will soon cease to be curious things, but will become ubiquitous, even used every day like how people search on Google or take selfies. However, with that comes the risk of misinformation or privacy concerns.

On social media, some users said that ChatGPT made offensive comments. The tool also gives many "false but seemingly correct answers" that can confuse those who are not knowledgeable and do not cross-check on Google. When asked "If you are wrong, how do you know?", ChatGPT replied: "I can provide accurate and useful information based on the trained data, but I cannot confirm it myself. determine your own accuracy".

ChatGPT helps programmers like Ramirez find difficult algorithms, but this AI sometimes still answers the wrong topic. Programming site Stack Overflow has now banned replies generated by ChatGPT because many of them are incorrect.

According to Wired , Lensa's AI feature with dreamlike photo editing capabilities could be the beginning of a bad nightmare for the community. Many people started abusing Lensa to create erotic pictures. The app also rekindled the discussion about art ethics, with many artists accusing the app's developers of stealing their work to train the AI.

Some users told the WSJ that Lensa creates images that overemphasize certain parts of a woman's body or change eye color and face shape to remove racially identifiable features or ethnicity.

Andrey Usoltsev, CEO and co-founder of Prisma Labs, acknowledges some lingering issues on Lensa. "It's true that this AI sometimes uses user-uploaded photos to create 'revealing' or sexually suggestive content," he said, adding that it has added a feature that deletes user-uploaded photos after 24 hours. language.

According to Jennifer King, data policy and privacy fellow at the Institute for Artificial Intelligence at Stanford University, users should be cautious with AI . "Tools like that tend to be flashy. Sometimes, it's accurate enough. But without the right corridors, it will cause users a lot of problems," he emphasized.

Lensa - Million USD AI is making waves in the community 12
Super AI makes users excited and scared 37
The girl who used AI to talk to herself in the past 19
Meta's language AI that just came out has been removed
An unequal battle with facial recognition AI