What is Character AI?

March 3, 2023

Top ChatGPT Alternatives to Try in 2023


As an AI language model, ChatGPT has become incredibly popular for its ability to generate human-like responses to a wide range of queries. It has become a go-to option for individuals and businesses looking to automate their communication processes. However, with the rapid advancements in AI technology, several ChatGPT alternatives have emerged in recent times. In this article, we’ll take a closer look at some of the best ChatGPT alternatives available on the market.

Without wasting much of your time, here are the top ChatGPT alternatives to try in 2023. This list is in no particular order and contains both free and paid AI chatbots.

Advertisement - Continue reading below


YouChat is a conversational AI platform that allows businesses to easily create and deploy conversational bots for messaging and voice services. It uses natural language processing (NLP) and machine learning to understand user intents and provide relevant responses.

Try YouChat here.

YouChat is a free ChatGPT-like AI search assistant that you can talk to right in your search results.

Read More: Here is How to Get Started With ChatGPT


Jasper AI is an innovative writing tool that utilizes advanced AI technology to create content at a rate five times faster than the average human copywriter. Jasper AI also includes pre-written templates for specific categories, enabling users to customize and tailor their content to their specific needs.

Jasper is not free and your first 10,000 words are free to try. If you choose to continue, plans start at $29/mo.

Try Jasper here.


ChatSonic is another incredibly powerful conversational AI chatbot by Writesonic. The advanced AI chatbot is based on the latest GPT-3.5 model and uses Natural Language Processing (NLP) and Machine Learning (ML) technology to automate the text and image generation process.

Advertisement - Continue reading below

Chatsonic AI offers a free trial for users to test the platform’s capabilities. However, the free trial is limited to 2,500 words.

You can get started here.

Related: ChatGPT Plus Vs ChatGPT: Main Difference and How to Upgrade

IBM Watson

IBM Watson is a remarkable computer system that can answer questions posed in natural language. It was developed by a research team led by principal investigator David Ferrucci as part of IBM’s DeepQA project.

The system is named after IBM’s founder and first CEO, industrialist Thomas J. Watson, and is known for its exceptional ability to understand and analyze complex language and data. You will need to create an account on IBM Cloud to try Watson services for free with no time restrictions. Read More.


Google Bard is a new chatbot tool that was recently launched as a competitor to other advanced artificial intelligence systems, such as ChatGPT. The platform is designed to simulate human-like conversations and employs a combination of natural language processing and machine learning to provide users with realistic and helpful responses to their queries.

Unfortunately, Google Bard has yet to be made available to the public. Follow this blog for more information.

Read More: Here is Google Bard: ChatGPT’s New Rival

Caktus AI

This is one of the leading artificial intelligence-powered platforms in education. Their goal is to help students transition from high school to a career. Caktus AI focuses on education rather than doing everything like ChatGPT.

Advertisement - Continue reading below

Caktus has a free trial of 5,000 characters—which is about 500-1,000 words. The Caktus Premium plan comes at $9.99/month.

Get started here.


According to the company, this chatbot program doesn’t just talk to people, it learns their texting styles to mimic them. It is dubbed as an AI companion that cares. Try this chatbot here.

Honourable Mentions

In addition to the popular AI chatbots we have already discussed, there are several other AI-powered chatbots worth mentioning. These chatbots offer a range of features and capabilities. They are designed to meet the needs of businesses across a range of industries

Socratic for Students

LUIS – L anguage U nderstanding I ntelligent S ervice.

anguage nderstanding ntelligent ervice. LaMDa by Google

DeepL Write


Character AI

What is the best AI Chatbot?

ChatGPT is a popular AI chatbot due to its impressive performance, versatility, and free availability. Powered by OpenAI’s GPT-3 language model, it excels in various language tasks, such as writing, summarization, translation, and conversation.

Its proficiency and user-friendly interface make it an ideal option for businesses and individuals seeking to automate their communication processes.


The list above contains several AI chatbot alternatives to ChatGPT, each with its own unique features and capabilities. We hope you will have the chance to try out some of these chatbots, and please let us know in the comment section about your experience.

Subscribe to read


What is included in my trial?

During your trial you will have complete digital access to FT.com with everything in both of our Standard Digital and Premium Digital packages.

Standard Digital includes access to a wealth of global news, analysis and expert opinion. Premium Digital includes access to our premier business column, Lex, as well as 15 curated newsletters covering key business themes with original, in-depth reporting. For a full comparison of Standard and Premium Digital, click here.

Change the plan you will roll onto at any time during your trial by visiting the “Settings & Account” section.

What happens at the end of my trial?

If you do nothing, you will be auto-enrolled in our premium digital monthly subscription plan and retain complete access for $69 per month.

For cost savings, you can change your plan at any time online in the “Settings & Account” section. If you’d like to retain your premium access and save 20%, you can opt to pay annually at the end of the trial.

You may also opt to downgrade to Standard Digital, a robust journalistic offering that fulfils many user’s needs. Compare Standard and Premium Digital here.

Any changes made can be done at any time and will become effective at the end of the trial period, allowing you to retain full access for 4 weeks, even if you downgrade or cancel.

Who will compete with ChatGPT? Meet the contenders | The AI Beat


Today, ChatGPT is two months old.

Yes, believe it or not, it was less than nine weeks ago that OpenAI launched what it simply described as an “early demo” a part of the GPT-3.5 series — an interactive, conversational model whose dialogue format “makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.”

ChatGPT quickly caught the imagination — and feverish excitement — of both the AI community and the general public. Since then, the tool’s possibilities as well as limitations and hidden dangers have been well established, and any hints of slowing down its development were quickly dashed when Microsoft announced its plans to invest billions more into OpenAI.

Can anyone catch up and compete with OpenAI and ChatGPT? Every day it seems like contenders, both new and old, step into the ring. Just this morning, for example, Reuters reported that Chinese internet search giant Baidu plans to launch an AI chatbot service similar to OpenAI’s ChatGPT in March.

Here are four top players potentially making moves to challenge ChatGPT:

Anthropic: Claude

According to a New York Times article last Friday, Anthropic, a San Francisco startup, is close to raising roughly $300 million in new funding, which could value the company at around $5 billion.

Keep in mind that Anthropic has always had money to burn: Founded in 2021 by several researchers who left OpenAI, it gained more attention last April when, after less than a year in existence, it suddenly announced a whopping $580 million in funding — which, it turns out, mostly came from Sam Bankman-Fried and the folks at FTX, the now-bankrupt cryptocurrency platform accused of fraud. There have been questions as to whether that money could be recovered by a bankruptcy court.

Anthropic, and FTX, has also been tied to the Effective Altruism movement, which former Google researcher Timnit Gebru called out recently in a Wired opinion piece as a “dangerous brand of AI safety.”

Anthropic developed an AI chatbot, Claude — available in closed beta through a Slack integration — that reports say is similar to ChatGPT and has even demonstrated improvements. Anthropic, which describes itself as “working to build reliable, interpretable, and steerable AI systems,” created Claude using a process called “Constitutional AI,” which it says is based on concepts such as beneficence, non-maleficence and autonomy.

According to an Anthropic paper detailing Constitutional AI, the process involves a supervised learning and a reinforcement learning phase: “As a result we are able to train a harmless but non-evasive AI assistant that engages with harmful queries by explaining its objections to them.”

DeepMind: Sparrow

In a TIME article two weeks ago, DeepMind’s CEO and cofounder Demis Hassabis said that DeepMind is is considering releasing its chatbot Sparrow in a “private beta” sometime in 2023. In the article, Hassabis said it is “right to be cautious” in its release, so that the company can work on reinforcement learning-based features like citing sources — something ChatGPT does not have.

DeepMind, which is the British-owned subsidiary of Google parent company Alphabet, introduced Sparrow in a paper in September. It was hailed as an important step toward creating safer, less-biased machine learning (ML) systems, thanks to its application of reinforcement learning based on input from human research participants for training.

DeepMind says Sparrow is a “dialogue agent that’s useful and reduces the risk of unsafe and inappropriate answers.” The agent is designed to “talk with a user, answer questions and search the internet using Google when it’s helpful to look up evidence to inform its responses.”

However, DeepMind has said it considers Sparrow a research-based, proof-of-concept model that is not ready to be deployed, according to Geoffrey Irving, a safety researcher at DeepMind and lead author of the paper introducing Sparrow.

“We have not deployed the system because we think that it has a lot of biases and flaws of other types,” Irving told VentureBeat last September. “I think the question is, how do you weigh the communication advantages — like communicating with humans — against the disadvantages? I tend to believe in the safety needs of talking to humans … I think it is a tool for that in the long run.”

Google: LaMDA

You might remember LaMDA from last summer’s “AI sentience” whirlwind, when Blake Lemoine, a Google engineer, was fired due to his claims that LaMDA — short for Language Model for Dialogue Applications — was sentient.

“I legitimately believe that LaMDA is a person,” Lemoine told Wired last June.

But LaMDA is still considered to be one of ChatGPT’s biggest competitors. Launched in 2021, Google said in a launch blog post that LaMDA’s conversational skills “have been years in the making.”

Like ChatGPT, LaMDA is built on Transformer, the neural network architecture that Google Research invented and open-sourced in 2017. The Transformer architecture “produces a model that can be trained to read many words (a sentence or paragraph, for example), pay attention to how those words relate to one another and then predict what words it thinks will come next.”

And like ChatGPT, LaMDA was trained on dialogue. According to Google, “During its training, [LaMDA] picked up on several of the nuances that distinguish open-ended conversation from other forms of language.”

A New York Times article from January 20 said that last month, Google founders Larry Page and Sergey Brin met with company executives to discuss ChatGPT, which could be a threat to Google’s $149 billion search business. In a statement, a Google spokeswoman said: “We continue to test our AI technology internally to make sure it’s helpful and safe, and we look forward to sharing more experiences externally soon.”

Character AI

What happens when engineers who developed Google’s LaMDA get sick of Big Tech bureaucracy and decide to strike out on their own?

Well, just three months ago, Noam Shazeer (who was also one of the authors of the original Transformer paper) and Daniel De Freitas launched Character AI, its new AI chatbot technology that allows users to chat and role-play with, well, anyone, living or dead — the tool can impersonate historical figures like Queen Elizabeth and William Shakespeare, for example, or fictional characters like Draco Malfoy.

According to The Information, Character “has told investors it wants to raise as much as $250 million in new funding, a striking price for a startup with a product still in beta.” Currently, the report said, the technology is free to use, and Character is “studying how users interact with it before committing to a specific plan to generate revenue.”

In October, Shazeer and De Freitas told the Washington Post that they left Google to “get this technology into as many hands as possible.”

“I thought, ‘Let’s build a product now that can that can help millions and billions of people,’” Shazeer said. “Especially in the age of COVID, there are just millions of people who are feeling isolated or lonely or need someone to talk to.”

And, as he told Bloomberg last month: “Startups can move faster and launch things.”

What is Character AI?


What is Character AI?

Ai generated characters? How interesting

Table of Contents

Table of Contents

Table of Contents hide Character creation

Character AI is an exciting new development in the world of artificial intelligence. Created by former Google LaMDA developers Noam Shazeer and Daniel De Freitas, this neural language model chatbot web application is capable of generating human-like text responses and engaging in contextual conversations.

Despite only being in beta mode, Character AI has already demonstrated outstanding results, bringing science-fiction ideas of open-ended dialogues and collaborative interactions with machines to life.

The technology behind Character AI involves complex algorithms and machine learning that enable the chatbot characters to behave more like real people.

The algorithms take into account various factors, including the character’s personality, emotions, and goals, as well as the environment they’re in and the actions of other characters around them. With Character AI, you can converse with anyone, from Elon Musk to Pikachu, and even talk to deities! This innovative technology opens up endless possibilities for communication and interaction, making it an exciting development in the world of AI.

Character AI is part of a larger trend toward the creation of more interactive and responsive technology.

As AI technology continues to evolve, character AI is becoming increasingly sophisticated and capable of mimicking human behavior. This has important implications for the future of human-computer interaction, as it could lead to more natural and intuitive communication with machines.

Moreover, character AI has the potential to revolutionize the entertainment industry, allowing for more immersive and engaging experiences in movies, TV shows, and video games.

As the technology continues to improve, we can expect to see more and more characters that are indistinguishable from their human counterparts, blurring the lines between reality and fantasy.

READ NOW: Chat GPT vs Notion AI

Character creation

The personalization of characters is what sets Character AI apart from other chatbots. Users have the ability to create characters that are tailored to their needs, preferences, and interests. The personality of each character is carefully crafted to reflect their traits, interests, and characteristics.

This is done by providing the AI with a series of descriptions from the point of view of the character and its greeting message, which are then molded from conversations to create examples. The AI uses these examples to modify the character’s messages to fit the desired dialect and identity.

Users can further refine the character’s personality and responses by providing feedback through the rating system. Another exciting aspect of Character AI is the ability to have group chats containing multiple characters.

This allows for unique interactions between characters and users, creating an immersive and dynamic experience. With the increasing conversation memory capacity, the AI can “remember” messages from farther back, allowing for more contextual and engaging conversations.

Users have the option to choose between Quick Creation and Advanced Creation modes for building their characters. Quick Creation is a straightforward process that involves clicking “Create a Character” under “Create” in the menu on the left side of the website’s homepage.

In contrast, Advanced Creation allows users to input short and long descriptions of their character, along with example chats that give the AI a better understanding of the character’s behavior. This mode also allows for more complex features such as setting a scene and building image-generating characters.

For those who need guidance, an official guide on perfecting character creation, called the Character Book, is available on the website. However, the developers have made it clear that pornographic content is not supported and is against their terms of service.

Best ChatGPT AI Alternatives You Need To Try in 2023


The AI platform will behave as a General AI chatbot if you’d like it to, but there are various personas it can assume, such as a philosopher, astrologer, or stand-up comedian, something which ChatGPT doesn’t offer at the moment. ChatSonic has other advantages over ChatGPT too. For one, it can create AI-generated images, and you’ll be able to create 100 per month for free with ChatSonic:

On top of this, ChatSonic is connected to the internet – which means it’s able to provide real-time, up-to-date answers, especially concerning events that take place after 2021, of which ChatGPT has very limited knowledge.

  1. OpenAI Playground Price: Free

Free Set Up: OpenAI account required (email & password)

OpenAI account required (email & password) Best ChatGPT AI alternative for: Customization & experimentation If you find yourself locked out of ChatGPT and you don’t want to get to grips with a ChatGPT alternative just yet, then use OpenAI’s “Playground” tool. OpenAI playground is very similar to ChatGPT on the surface and is effectively just a more experimental version of the same tool – it’s actually been open for public use for longer than its more finely-tuned cousin. It’s a lot more customizable on the whole – you can even select which language model you’d like it to use before you ask it any questions. OpenAI says that Davinci-003, which we used for the tests below, often answers questions with “higher quality, longer output, and better instruction following.”

As you can see from the image above, the OpenAI Playground user interface isn’t quite as straightforward as ChatGPT’s, so you might find it a little more difficult to use – but that’s the price you pay for the extra customization options. When we’ve used OpenAI Playground, we’ve found that generally, it’s not as hesitant as ChatGPT when it comes to weighing in on complex, controversial topics, so you may get some interesting and unexpected answers while using it. However, it also, like ChatGPT, struggles with queries relating to events from after 2021.

  1. YouChat Price: Free

Free Setup? N/A (immediate usage – sign in send replies)

N/A (immediate usage – sign in send replies) Best ChatGPT AI alternative for? Chatbot + search engine combo YouChat is an AI chatbot and search assistant. It will provide you with similar answers to that of ChatGPT, but will also link you to webpages that relate to whatever query you choose to type in:

One advantage of YouChat is that you don’t even have to make an account, unlike ChatSonic and ChatGPT. Once you get onto the site, you can start using their AI Chatbot immediately. However, you will have to make an account if you want to continue your conversation with YouChat past its answer to your query. You can certainly have conversations with YouChat, but much like ChatGPT, it’s better for condensing down complex information found on the internet and searching for information in a more efficient, quicker way. But unlike ChatGPT, YouChat will also show you an article written by a human, so it’s the best of both worlds if that’s your primary use case. However, we did find YouChat slightly glitchier than ChatGPT and it decided to just not reply to a couple of queries we inputted into its search bar. Bear in mind, however, that it has only been fully operational since December 2022, so as is the case with ChatGPT, it’s likely to be further refined and improved.

  1. Perplexity AI Price from: Free

Free Set Up: N/A (immediate usage)

N/A (immediate usage) Best ChatGPT AI alternative for: Citing sources used to generate responses Perplexity functions quite similarly to a search engine. However, Perplexity will produce an answer that combines information from top web results, rather than just providing you with a list like Google does. It’s not quite as big of a project as some of the other AI chatbots on this list, maintained by just a small, 8-person team – but it’s still an impressive ChatGPT AI alternative. One thing that sets Perplexity apart is the way it neatly collates the sources used to create the answers it serves to users, which means you can judge whether they’re legitimate and credible. You don’t have this power with ChatGPT, and in a world crippled by misinformation, it’s an excellent feature. ChatSonic also has this feature, but it doesn’t cite its sources every time as Perplexity does. While using Perplexity, it was nice to have a list of related queries listed below every answer. An answer to any of the suggested queries can be generated instantly with a single click.

However, Perplexity did take slightly longer to produce an answer than some of the other entries to this list. Now that could be down to various factors (from site traffic to the tester’s internet connection) – but all the other chatbots on this list produced answers quicker.

  1. Character AI Price from: Free

Free Set Up: N/A (immediate usage)

N/A (immediate usage) Best ChatGPT AI alternative for: Entertaining responses Character AI does what it says on the tin – it’s an artificial intelligence program that allows users to converse with artificially constructed simulations of famous characters from movies and TV shows, as well as public figures and celebrities. One neat feature of Character AI is you can build your own character bot very easily. All you have to do is come up with what you want your chatbot to do and set a few parameters – after that, Character AI will leaf through huge troves of literature and written text to ensure the AI bot sounds like you intend it to. Character AI does have its downsides. When we tested it, we found that some responses weren’t necessarily “out of character”, but weren’t particularly “in character” either – although these Super Mario and Elon Musk simulations certainly were:

Plus, its answers are definitely not as consistently accurate or as useful as the ChatGPT – it’s a bit more of a novelty tool to have a bit of fun with. If you’re just looking for an AI program to entertain you, then there aren’t many better than Character AI.

  1. Jasper Chat Price from: 5-day free trial, $29/month afterward (Starter plan)

5-day free trial, $29/month afterward (Starter plan) Set Up: Account required + answers to intro questions

Account required + answers to intro questions Best ChatGPT AI alternative for: Content Generation Jasper Chat is a powerful AI tool that can help advertising and marketing businesses producing online content scale their operations. It’s effectively an AI content-generating robot that has significantly more business-focused features than the other ChatGPT alternatives on this list. It can write adverts, create captions for social media posts, produce video scripts, and much more.

Jasper Chat will also remember past queries, conversations, and prompts that you type into its interface and factor these into how it responds, which ChatGPT doesn’t have the capacity to do. However, it’s similar to ChatGPT in the sense that it struggles with information relating to events that took place post-2021. Jasper has a $29 per month Starter plan, but if you purchase the $49 per month, however, Jasper Chat will generate 50,000 words, while features in the package include Compose & Command features, as well as a Google docs style editor.

Why You Should Look for a ChatGPT Alternative ChatGPT is a fantastic resource for individuals and businesses, but as with every type of software, no matter how useful, it has some downsides. Along with the ChatGPT alternatives featured in this article, Tech.co has also been experimenting with ChatGPT and tracking developments since it came out to see exactly what it’s capable of. During our testing, we’ve identified four major downsides: The site is often at capacity. Due to its soaring popularity, sometimes, ChatGPT has so many visitors it becomes inaccessible to users who want to start conversations.

Due to its soaring popularity, sometimes, ChatGPT has so many visitors it becomes inaccessible to users who want to start conversations. It has limited knowledge of post-2021 events. ChatGPT is not well-trained to respond to questions about events that happened after 2021. This makes its answers to certain topics unintelligible or useless.

ChatGPT is not well-trained to respond to questions about events that happened after 2021. This makes its answers to certain topics unintelligible or useless. It is expensive to run and will quickly be monetized. ChatGPT is unlikely to be free forever, at least in its most advanced form – it was revealed recently that it costs “single digit cents per chat” per conversation.

ChatGPT is unlikely to be free forever, at least in its most advanced form – it was revealed recently that it costs “single digit cents per chat” per conversation. It has only trained on text datasets. ChatGPT can hold a conversation, but it’s training data rather than human conversations, so it doesn’t “feel” human and doesn’t converse like one.

AI Updates: Bing, Bard and Chinchilla While this article focuses on ChatGPT AI alternatives, here’s a quick rundown of other AI projects and dvelopments to be aware of.

Microsoft Bing with ChatGPT Microsoft recently launched a new version of Bing with ChatGPT technology integrated into the search engine for more accurate, complex searches. However, currently, it’s only available as a limited preview, which you can access by heading over to the Bing landing page. You’ll be able to see how it works for a few basic queries, but you won’t be able to use it like ChatGPT yet. However, it’s expected that users will soon be able to sign up for the full version.

Introducing the AI Mirror Test, which very smart people keep failing


In behavioral psychology, the mirror test is designed to discover animals’ capacity for self-awareness. There are a few variations of the test, but the essence is always the same: do animals recognize themselves in the mirror or think it’s another being altogether?

Right now, humanity is being presented with its own mirror test thanks to the expanding capabilities of AI — and a lot of otherwise smart people are failing it.

Don’t be distracted by what you see in the mirror. Image: Xavier Hubert-Brierre via Tenor

The mirror is the latest breed of AI chatbots, of which Microsoft’s Bing is the most prominent example. The reflection is humanity’s wealth of language and writing, which has been strained into these models and is now reflected back to us. We’re convinced these tools might be the superintelligent machines from our stories because, in part, they’re trained on those same tales. Knowing this, we should be able to recognize ourselves in our new machine mirrors, but instead, it seems like more than a few people are convinced they’ve spotted another form of life.

This misconception is spreading with varying degrees of conviction. It’s been energized by a number of influential tech writers who have waxed lyrical about late nights spent chatting with Bing. They aver that the bot is not sentient, of course, but note, all the same, that there’s something else going on — that its conversation changed something in their hearts.

“No, I don’t think that Sydney is sentient, but for reasons that are hard to explain, I feel like I have crossed the Rubicon,” wrote Ben Thompson in his Stratechery newsletter.

“In the light of day, I know that Sydney is not sentient [but] for a few hours Tuesday night, I felt a strange new emotion — a foreboding feeling that AI had crossed a threshold, and that the world would never be the same,” wrote Kevin Roose for The New York Times.

In both cases, the ambiguity of the writers’ viewpoints (they want to believe) is captured better in their longform write-ups. The Times reproduces Roose’s entire two-hour-plus back-and-forth with Bing as if the transcript was a document of first contact. The original headline of the piece was “Bing’s AI Chat Reveals Its Feelings: ‘I Want to Be Alive” (now changed to the less dramatic “Bing’s AI Chat: ‘I Want to Be Alive.’”), while Thompson’s piece is similarly peppered with anthropomorphism (he uses female pronouns for Bing because “well, the personality seemed to be of a certain type of person I might have encountered before”). He prepares readers for a revelation, warning he will “sound crazy” when he describes “the most surprising and mind-blowing computer experience of my life today.”

Having spent a lot of time with these chatbots, I recognize these reactions. But I also think they’re overblown and tilt us dangerously toward a false equivalence of software and sentience. In other words: they fail the AI mirror test.

What is important to remember is that chatbots are autocomplete tools. They’re systems trained on huge datasets of human text scraped from the web: on personal blogs, sci-fi short stories, forum discussions, movie reviews, social media diatribes, forgotten poems, antiquated textbooks, endless song lyrics, manifestos, journals, and more besides. These machines analyze this inventive, entertaining, motley aggregate and then try to recreate it. They are undeniably good at it and getting better, but mimicking speech does not make a computer sentient.

“What I had not realized is that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people.”

This is not a new problem, of course. The original AI intelligence test, the Turing test, is a simple measure of whether a computer can fool a human into thinking it’s real through conversation. An early chatbot from the 1960s named ELIZA captivated users even though it could only repeat a few stock phrases, leading to what researchers call the “ELIZA effect” — or the tendency to anthropomorphize machines that mimic human behavior. ELIZA designer Joseph Weizenbaum observed: “What I had not realized is that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people.”

Now, though, these computer programs are no longer relatively simple and have been designed in a way that encourages such delusions. In a blog post responding to reports of Bing’s “unhinged” conversations, Microsoft cautioned that the system “tries to respond or reflect in the tone in which it is being asked to provide responses.” It is a mimic trained on unfathomably vast stores of human text — an autocomplete that follows our lead. As noted in “Stochastic Parrots,” the famous paper critiquing AI language models that led to Google firing two of its ethical AI researchers, “coherence is in the eye of the beholder.”

Researchers have even found that this trait increases as AI language models get bigger and more complex. Researchers at startup Anthropic — itself founded by former OpenAI employees — tested various AI language models for their degree of “sycophancy,” or tendency to agree with users’ stated beliefs, and discovered that “larger LMs are more likely to answer questions in ways that create echo chambers by repeating back a dialog user’s preferred answer.” They note that one explanation for this is that such systems are trained on conversations scraped from platforms like Reddit, where users tend to chat back and forth in like-minded groups.

ELIZA captivated users, even though it was only capable of repeating stock phrases. Image: Marcin Wichary / CC 2.0

Add to this our culture’s obsession with intelligent machines and you can see why more and more people are convinced these chatbots are more than simple software. Last year, an engineer at Google, Blake Lemoine, claimed that the company’s own language model LaMDA was sentient (Google said the claim was “wholly unfounded”), and just this week, users of a chatbot app named Replika have mourned the loss of their AI companion after its ability to conduct erotic and romantic roleplay was removed. As Motherboard reported, many users were “devastated” by the change, having spent years building relationships with the bot. In all these cases, there is a deep sense of emotional attachment — late-night conversations with AI buoyed by fantasy in a world where so much feeling is channeled through chat boxes.

To say that we’re failing the AI mirror test is not to deny the fluency of these tools or their potential power. I’ve written before about “capability overhang” — the concept that AI systems are more powerful than we know — and have felt similarly to Thompson and Roose during my own conversations with Bing. It is undeniably fun to talk to chatbots — to draw out different “personalities,” test the limits of their knowledge, and uncover hidden functions. Chatbots present puzzles that can be solved with words, and so, naturally, they fascinate writers. Talking with bots and letting yourself believe in their incipient consciousness becomes a live-action roleplay: an augmented reality game where the companies and characters are real, and you’re in the thick of it.

But in a time of AI hype, it’s dangerous to encourage such illusions. It benefits no one: not the people building these systems nor their end users. What we know for certain is that Bing, ChatGPT, and other language models are not sentient, and neither are they reliable sources of information. They make things up and echo the beliefs we present them with. To give them the mantle of sentience — even semi-sentience — means bestowing them with undeserved authority — over both our emotions and the facts with which we understand in the world.

< 回到列表