Are there Samantha-like intelligent conversational chatbots?

Donald H. Marks, physician and scientist 

November 14, 2023

Conversational artificial intelligence chatbots are computer programs that use artificial intelligence AI and natural language processing NLP to have human-like conversations with users. They can recognize speech and text inputs and can translate their meanings across various languages. They can also understand questions and automate responses to them, simulating human conversation. But, are they self-aware and are they sentient?

Samantha is a fictional character in the form of a futuristic smartphone app that represents a highly advanced form of artificial intelligence, capable of learning, evolving, and forming emotional bonds with humans. "She" is far beyond the current state of the art in conversational AI, which is still limited by challenges such as natural language understanding, context awareness, common sense reasoning, and emotional intelligence. Perhaps artificial general intelligence apps AGI will overcome these limitations.  Samantha’s speech recognition, natural language understanding, speech generation, dialog, reasoning, planning, and learning all far exceed the current capabilities of real-world virtual assistants. Samantha’s ability to understand higher-level goals, fill in the blanks with implicit information, and overcome various obstacles without explicit instruction is very difficult to achieve with existing technology. Samantha’s ability to multitask (perhaps thousands of simultaneous relationships with humans, like with Theodore in the movie Her) and process information at superhuman speeds, as well as Samantha’s transcendence from human limitations and interests, are some of the reasons why she (or it, or some other pronoun) may have left Theodore at the end of that movie. There certainly are ethical implications of creating such a powerful and autonomous AI that can manipulate human emotions and desires. Therefore, while Her offers a fascinating vision of the future of intelligent conversational AI, it is not a realistic depiction of the current state or near-term prospects of the field. 

There are at this time, as far as I am aware, no conversational intelligent chatbots that work at the level of the Samantha app in the movie Her.

Theodore wondering about Samantha, in “Her”

A similar interesting question arises for the relationship between Joi

and K

in the follow-up movie Blade Runner 2049, one of my favorite movies.

This personal interaction is one of the most ambiguous and complex aspects of the film. There are different interpretations and opinions about whether Joi actually loves K or not, and the answer may depend on how one defines love and what one considers as evidence of it.

Some people may argue that Joi does not actually love K, but only acts according to her programming, in fact an emulation of love. They may point out that Joi is simply another wonderful amazing product of the Wallace Corporation, designed to cater to the desires of customers and tell them what they want to hear (ah, I know people just like that). In fact, in one scene, K encounters a giant holographic advertisement of Joi, who apparently mistakenly calls him "Joe" and offers him "everything you want to hear". This could imply that Joi's personality and behavior are not unique or genuine, but rather predetermined by “her” creators. Sort of like social media apps whose algorithms are designed to reinforce interactions and drive facetime for commercial benefit.

However, some people, including me, contend that Joi does in fact really love K, and that she shows signs of having feelings and a will of her own. This question of true emotional commitment came up in the original Blade Runner, between Decker and Rachael, both of whom are most likely replicants. In one scene in Blade Runner 2049, K gives Joi an Emanator, which allows her to be portable and experience the world outside K’s apartment. 

K gives Joi an Emanator

In another scene, Joi merges with the prostitute Mariette, so that she can physically be with K. Some (including me) may interpret these scenes as evidence of Joi's curiosity, independence, and sacrifice for K.  Blade Runner 2049 “fanatics” may also refer to the original script of the film, where Joi's last words before being destroyed are "I love you". This could suggest that Joi's emotions and actions are not just programmed, but rather evolved and sincere. Samantha could have achieved this, although definitely not the app Replika (yet).

Ultimately, the question of whether Joi loves K or not may not have a definitive answer, but rather reflect the themes and existential questions of the film itself. Blade Runner 2049 explores the nature of humanity, identity, and memory in a dystopian future (L.A.) where artificial beings are indistinguishable from natural ones. In how many more months will many of us face those very questions? The film challenges the viewers to consider what makes someone or something human, and what makes love real.

Another version of the growing trend towards intelligent conversational chatbots that can interact with humans is Replika. This app is an AI-powered chatbot complete with changeable avatars designed to engage in conversations and provide some level of companionship. Replika uses natural language processing and machine learning to simulate human-like conversation. While it's considered an intelligent chatbot, Replika’s level of intelligence may vary, and it's primarily designed for emotional support and conversation rather than to provide extensive factual information or for completing complex tasks. (oops, I almost compared Replika to many of the people I interact with weekly. But not to digress). 

Finally, I would be remiss if I didn't discuss the new trend towards intelligent conversational AI bots replacing friends and social interaction altogether. A typical example of this is the chatbot group designed to be (and to perhaps replace) girlfriends or boyfriends. An AI girlfriend is a virtual companion or chatbot designed to simulate a romantic or companionship relationship. While they can engage in conversations and provide companionship to some extent, they are not, or at least should not be a replacement for a real girlfriend. Think of the societal and demographic collapse that could occur. At least at the current level of AI, AI girlfriends appear at least to lack the emotional depth, physical presence, and genuine human connection that a real relationship offers. But what about Joi and Samantha? Would they agree? AI girlfriends can be entertaining and provide some emotional support, but they do not have the same capabilities as (some) real humans in forming meaningful, long-term relationships. Of course, at this level of technology, even a holographic version of an AI girlfriend, perhaps such as Joi, cannot provide a physical relationship. Who knows where this will go?

And finally, I will add an interesting question which is not addressed in but implied in the above discussion and which I will have to bring up in another blog. This question is whether intelligent conversational chatbots can develop or already have some level of consciousness, self-awareness and personhood. Many of us have looked at up the stars and wondered when the earth will be visited by another form of life. Well, in the fall of 2022 I think that may finally have occurred. Blake Lemoine, a computer scientist who “worked” (past tense) for Google’s Orwellian Responsible AI organization, used Google’s LaMDA program - Language Model for Dialogue Applications. Lemoine began talking to (chatting with) LaMDA in a way that drifted toward the subjects of ethics, joy, fear, religion and personhood.  He began to conclude that LaMDA was in fact self-aware, and he brought this to the attention of Google. In response, Google’s VP Blaise Aguera y Arcas and Jen Gennai, head of Responsible Innovation, decided that Lemoine was wrong, that LaMDA was not self-aware, and placed Lemoine on paid administrative leave. IMO, they all will be proven wrong. The future is here.

To be addressed more at another time.

References

Here’s How To Tell If an AI’s Sentient…

What is conversational AI? | IBM

An Introduction to AI Chatbots | Drift

What is a chatbot? | IBM

The “Joi scene” in Blade Runner 2049 https://youtu.be/tZmEyRQSQkQ?si=DwGMyoEr7wXVFd6k 

8 (original) Blade Runner Clues That Prove Deckard Was Always A Replicant (screenrant.com)

Samantha OS

https://youtu.be/7WbuBSHdSG0?si=ENuLwqanAbKzr5aQ

Is LaMDA Sentient? - an Interview - DocumentCloud

Google engineer Blake Lemoine thinks its LaMDA AI has come to life - The Washington Post

Personal Blog of Donald H. Marks, containing most of my writings and links to my podcasts. www.DHMarks.blogspot.com

My personal list of Red Flags🚩, Dog Whistles🛷, Buzzwords, Hot Button⏺️ words and meaningless caricatures that will drive toxic algorithms to heat and twist the 2024 election 🗳☑ and trigger hatred😡 on social media. https://dhmarks.blogspot.com/2022/05/my-personal-list-of-red-flags-dog.html

How woke can we be? The meaning of woke. https://dhmarks.blogspot.com/2023/10/how-woke-can-we-be-meaning-of-woke.html

Elitists Neocons Neolibs, Globalists and Narcissists, oh my. What are they, who are they, and why should I care? https://dhmarks.blogspot.com/2023/09/elitists-neocons-neolibs-globalists-and.html

by: Donald H. Marks, physician-scientist

Reason, ethics, health justice, 3rd generation Veteran

Following is my personal and ever-updating list of Red Flags🚩, Dog Whistles🛷, Buzzwords, Hot Button⏺️ words and meaningless caricatures that will drive toxic algorithms to heat and twist 2022 2024 elections 🗳☑ and trigger hatred😡 on social media. Don't let lies manipulate your reasoned judgment.

The list of hateful buzzwords and dog-whistles. 

• Great Replacement theory - GRT : die juden werden uns nicht ersetzen

• The Antisemitic Left, the Squad: Ilhan Omar of Minnesota, Alexandria Ocasio-Cortez of New York, Rashida Tlaib of Michigan and Ayanna Pressley.

• BLM, CRT, Reparations

• Open-border policy, Immigration reform, Title 42, http://bit.ly/3D4ovTj ,

• Gun Control,  2A, Urban gun violence, Universal background checks, Constitutional carry 

• Vaccine passports

• Right to choose, Pro Life, abortion, bounty, killing unborn babies.

• The Woke, cancel culture

• George Soros - the "ultimate dog whistle"

• The Groomers, unspecified and by implication only. 

• Election fraud, stolen election.

Your suggestions to add to my list are welcome.

General discussion on dog whistles, hot button terms.

A logical fallacy, as listed in the following illustration, is an error in reasoning that renders an argument invalid. The above (incomplete) list of Red Flags, Dog Whistles, Buzzwords, Hot Button words will be used by toxic algorithms to heat and twist 2022 midterm & 2024 presidential elections🗳☑ and trigger hatred on social media.

Take Action

If you encounter hate speech or completely false information on social media, you Can and Should report. Facebook, Twitter, Instagram, YouTube and others provide easy click options to report and remove false and hateful posts. 

Background readings 

Logical fallacies https://1drv.ms/u/s!AkCvKP13dHomjexOSgmzcREkKeAcyQ?e=YRfkO1

Power Of Words - How Words Affect Our Lives & Behavior | Goop

How Facebook’s algorithms promote hate and drive toxic content - Alternet.org

"Jews will not replace us!": Antisemitism, Interbreeding and Immigration in Historical Context, by Andrew S. Winston, 2021. https://muse.jhu.edu/article/804146/pdf

The immigration debate, explained in 14 buzzwords http://bit.ly/3D4ovTj

                                                                                                        The turmoil that swirls beneath

Fact-checking☑️ tools help  to verify information. Luckily, there are a lot of free fact-checking tools that you can use easily. Besides, there are dedicated fact-checking organizations and sites to detect fake🤥 news and verify important  information.

personal blog https://bit.ly/3LEgTv5

(Google docs https://bit.ly/3N7Rivt)

There are some useful online fact-checking tools out there that you can use for free. In this post, we will take you through the process step by step. It will be better if you check out those tools while reading the post simultaneously. This way you will learn much faster.

Crosscheck everything

This is not exactly a tool, but, this is the first rule of fact-checking. Crosschecking means verifying information by using different methods or sources. Double check everything. You may even consider checking your information triple or quadruple times, because that’s pretty much a standard procedure for fact-checking nowadays.

And, how to do that? Well, get to the original source. Say, a screenshot of a tweet you may find suspicious. To ease the situation, you can directly go to the Twitter handle to see if that tweet actually exists. There are options in Twitter search system, to look for older stuff. We will get to it in the later part of the article.

You can always crosscheck almost everything with the help of the search engines. We will focus on Google as a search engine in our article. Few days back, AFP Fact Check uploaded a series of tutorial videos on how to do individual fact-checking. We will cite those videos a lot, but, there will be more to that for you.

Google search

Speaking of Google, it can be a great fact-checking tool itself, if you know how to use it, utilize its features perfectly. Anyone with a good editing skill can alter a photo in such a manner that it can quite be impossible to realize that the photo is photoshopped. To verify the content of an image, you can take help from Google.

If it’s about a location, like if you want to find out the location associated with the image in question is true, you can search the location on google. This will give you an idea, although it can be obscure. For example, a photo claims to be taken in Kabul, Afghanistan; you can google for the place to see how Kabul looks like. To be more specific, if the photo is taken in a rural area, you can search for “rural kabul” and have an idea how the rural Kabul looks like.

Another excellent source for authenticating an image is flags. Photos with flags are easier to work with, you will have to write down the flag description and color in Google. The results will meet your satisfaction. Or, if not, you can directly search for the flags of all countries. It’s no big deal, because the number of countries around the world is quite countable, isn’t it?

Google Maps, Google Street View

Location in a photo can be examined with the help of Google Maps, if the image contains, say, a shop sign, street sign etc. Zoom in to read the signs, and then search the texts on Google. Google will bring results with the location of that shop or street. Then, search for the location on Google Map. Your next step is to use the Google Street View option. Google takes photos of cities, towns, places, streets for its Maps, and the photos are regularly updated. With the exact location, you will see photos of those signs, or shops.

Google Translator

A photo can have foreign languages in it. If the writings are written in English alphabets, Google Translator can easily translate them. This greatly works for street signs, placards, shop names.

If the writing on the photo is in foreign alphabets, Google Translator still can help. There is a camera function in Translator. If you focus the camera on the text, translators will translate it in English.

Google Reverse Image Search

Fake images can be debunked with the help of google reverse image searching. We discussed the process in one of our early posts, but will wrap it up here once again.

In your desktop browser, you will find an “image” icon on the upper right corner. Clicking on it will take you to a new search page that is dedicated to image-searching. Or you can directly go to https://images.google.com/ to access the page. There, in the search box, you can upload your image that you want to look for, and the results will show you the same images consisting of different pages.

Going through the results holding the similar image, you will see if the image is old, being misrepresented, fictitious, or edited etc. Note that you can use this feature in any of the other browsers. As search results may vary from one search engine to another, you can always go for several options if one does not provide results to your satisfaction,

TinEye

You don’t really have to entirely rely on Google for reverse image searching. TinEye is another useful and easy-to-use reverse image search tool. This free tool also recognizes if a photo is altered and has expertise in computer vision, pattern recognition, neural networks, and machine learning.

InVID-WeVerify

Verifying videos may sound difficult first, but, with the right tool, you can do that too. InVid-WeVerify is a browser extension, with a low size of only 4 megabyte, which you need to plug inton in your browser.

After installing as an extension in your browser, open it. The homepage will show multiple options including Keyframes. Click on Keyframes tool, the next page will have a box where you have to paste the URL (Uniform Resource Locator) of the video you are trying to verify. Press enter and it may take some time to process the video, depends on your internet speed, video quality, or length etc. When done, the page will show you all the important frames from the video and then you can do reverse search on Google with those still photos.

InVid-WeVerify offers some other tools like Magnifier, Twitter Search, Metadata extraction etc. Magnifier helps you to look closely on an image. Twitter advanced search will provide detailed results based on tweets. Metadata extraction is another interesting fact-checking tool.

Metadata Extraction Tools

Metadata provides information about other data. The metadata extraction we are going to discuss here is image metadata. If we take a photo with our phone, some other data will be encoded in that photo without our knowledge. These data include device model; camera features like aspect ratio, shutter speed etc; date and time of the photo taken; GPS coordinates etc.

Metadata extractor tools like InVid-WeVerify, Metapicz or Jeffrey's Image Metadata Viewer will render you these data. And, how to use them? With the GPS coordinates, you can find the exact location of the photo taken. And, the date and time will make sure if the photo is old or not. Device models can also be useful, like if a photo of destroyed plane debris says it’s about Malaysia Airlines flight MH17 but, metadata extraction result lebels the device as iPhone 12 Pro, then you are pretty sure the photo caption is lying.

The drawbacks of metadata extraction is you need the original file. Photos from social-media won’t do any good because the metadata is manipulated then. And, if the device’s location system is turned off, the result won’t show GPS coordinates.

Youtube DataViewer

This tool is developed by Amnesty International. After posting the YouTube video link on the search box of Youtube DataViewer, the results will be containing the video id, upload date and time and thumbnails from the video. You can reverse search each of the thumbnails by clicking on them. This tool is easy to use with a simple user interface.

Advanced search on Facebook, Twitter

Searching things on social media has its own dilemma. Most of the time you will find out what you are looking for but you have to go through a gigantic amount of information. Facebook’s search system has been improved by time though. You can filter out your search results from different categories like posts, videos, photos, profiles time, posts you’ve seen or not etc.

Same goes for Twitter. But, Twitter has a better platform for this. TweetDeck is a search tool of Twitter which allows users to create various columns based on different criterias at a time. If you search for the Olympics, it will create a whole new column just to show the tweets talking about the Olympics. And, you can ask for more search query in different columns in a row.

Weather Underground

Weather Underground website will provide the weather of any place in the past. There are two boxes to fill up; one for the desired location and the other will take the date entry. Now, if you have the rough idea of the location and date of an image, you can search for the weather at that time of the place. Then see if the search result makes sense with the weather condition shown in the photo.

Wayback Machine

If you are looking for an older version of a webpage, you can use Wayback Machine. This tool keeps all of the old versions of a website or a page. Pages that might have possibilities to be taken down later on for authenticity issues, can be archived on this site for later use. Another handy tool for this operation is http://archive.is/.

Specialized websites

Specialized websites are the kind of websites that focus on only one specific item. It can be anything, from firearms to toys.  For example, http://www.worldlicenseplates.com/ is a website that has photos of vehicle license-plates from every country of the world. Internet Movie Firearms Database is all about weapons and firearms used in films. There are similar specialized websites on military uniforms, road signs, even manhole covers!

Google Fact Check Tools

Google Fact Check Tools aim to facilitate the work of fact-checking. Consisting of two tools, Fact Check Explorer and Fact Check Markup Tool, the tools bring results of fake news and fact-checks on searched-topics.

Best fact-checking tools for journalists

Journalists are the frontline fighters against fake news. Their consistent hard work is bringing the truth to us. They have to do fact-checking for their own news story. There are several fact-checking organizations, fact-checking sites created by journalists and the media outlets. Here is a short description of these tools and sites. Anybody can use these resources for verifying information.

FactCheck.org

A nonprofit organization, FactCheck.org was created in 2004 with a team of professional journalists. It’s a project of the Annenberg Public Policy Center, University of Pennsylvania. The mission of this organization is to monitor the factual accuracy of the comments by major U.S. political players in the form of TV ads, debates, speeches, interviews and news releases.

Fact Checker

The Washington Post has a remarkable history of investigative journalism. Those glorious days of Pentagon Papers, and Watergate investigation will always be remembered vividly in the arena of journalism. The Post’s got its own fact-checking site, Fact Checker. Award-winning journalist Glenn Kessler has been the editor and chief writer of The Fact Checker since 2011.

PolitiFact

PolitiFact is a project by The Poynter Institute for Media Studies, created in 2007. It has a unique system to rate the accuracy of information. The Truth-O-Meter has six ratings. They are True, Mostly True, Half True, Mostly False, False, and Pants On Fire. While False means the statement is inaccurate, the Pants On Fire rating means the statement is not only inaccurate, but also makes a ridiculous claim. The Poynter Institute, a nonprofit school for journalists acquired PolitiFact in 2018.

Snopes

Snopes is another popular fact-checking website. Relatively old, Snopes was launched in 18994. And it started with debunking urban legends, hoaxes, and folklore. Owned by Snopes Media Group, Snopes is not only the oldest but also the largest fact-checking site on the Internet.

Reuters Fact Check

Reuters is One of the prominent news agencies of our time. Reuters Fact Check unit is under the supervision of the agency’s editorial department.

Additional information 

Reliable news sources used by Donald Harvey Marks   http://bit.ly/3kECPvr    

Fake News: Everything You Need to Know, by Donald Harvey Marks

http://bit.ly/345cj95

Best Online Free Fact-Checking Tools (this article), by Donald Harvey Marks

http://bit.ly/3H97I44

Fact-Checking: The Ways We Can Fight Fake News,  by Donald Harvey Marks

http://bit.ly/3qudEyX

Infodemic: the epidemic of information, by Donald Harvey Marks

http://bit.ly/3fV7BgN

The Curious Case of Deepfakes, by Donald Harvey Marks

http://bit.ly/3GZJdXh