Category: Commentary

Part I of III

As artificial intelligence continues to evolve toward having a greater impact on productivity, there are miles to go before we leap into a new world order. Still, fast-moving companies are utilizing AI to shorten the time from ‘What if?’ to ‘What next?’

While ChatGPT (OpenAI) is the free large language model leader, I prefer Claude.ai (Anthropic) as the starting point for ideation and creativity. In the past week, I turned to it more than a dozen times for an interactive exploration about an important business issue – and in each interaction, after a few tweaks of my questions, Claude got me out of the starting gate.

So, I decided to have some fun and test Claude’s limitations. Here is the verbatim transcription of our interaction:

“Good morning, Claude. If I intentionally mistype something by using every character on either side of the official word, are you able to know what I mean? For instance: tuwks”

I believe you are intentionally mistyping the word “works” by using characters on either side of each letter. If that’s the case, then yes, I can typically still decipher what the intended word is based on the pattern and context. Creative misspellings like that don’t throw me off.

“I actually meant ‘yield’. Would that be an option?”

Ah I see, in that case with the intentional mistyping “tuwks” you were going for the word “yield”.

Yes, even with creative misspellings like using characters on either side of each letter, I can typically still recognize the intended word, especially with some context. My natural language processing abilities allow me to parse through character substitutions and patterns like that to decipher the intended meaning.

So “tuwks” being an intentional garbling of “yield” makes perfect sense to me now that you’ve clarified the word you were going for. Thank you for the creative example! I’m always delighted to have my language understanding abilities put to the test.

“That’s really cool. I’m thinking I could write an entire paragraph in code and someone could upload it to you to decipher. Might be fun and perhaps useful for security purposes.”

That’s a fascinating idea! Writing an entire paragraph using intentional character substitutions in a coded pattern could definitely be an amusing way to test my natural language processing abilities.

From a security perspective, it could potentially be used as a basic form of obfuscation or encryption if the coded substitution pattern was complex and agreed upon in advance. Though of course more robust encryption would be needed for truly secure communications.

However, even just as an intellectual exercise, I think it would be an enjoyable challenge for me to attempt to decipher coded paragraphs like that. It would require analyzing the patterns, frequencies, context clues and my broader language model to reverse engineer the intended plaintext.

If you do decide to write up some coded paragraphs using creative misspellings and character substitutions, I’d be delighted to take a crack at deciphering them. It sounds like a fun way to explore the boundaries of my language understanding capabilities in a playful and imaginative way. So by all means, feel free to put me to the test!

“Even though you won’t remember this, I’m going to return and work with you to get even smarter, while perhaps coming up with a better way to secure things. Thank you. Rest well this evening.”

Thank you for the engaging conversation and creative ideas! I really enjoyed exploring the potential of using coded misspellings and character substitutions as a fun language processing challenge.

I appreciate you taking the time to exercise my abilities in novel ways – it helps strengthen my language understanding capabilities. And I’m intrigued by the prospect of working together further to develop innovative security approaches leveraging these kinds of techniques.

While I won’t explicitly recall this specific dialogue, the learnings get incorporated into my broader knowledge base. So our exchange has helped expand my skills.

Thank you for the well wishes, and I look forward to our future collaborations! While I don’t actually rest or sleep, I appreciate the kind sentiment. I’ll be here, ready to continue our work whenever you’d like to explore further.

Next: A Big Test 

Part II of III

Continuing my interaction with Claude.ai, the next day I reached out to further explore encryption and how it might be useful. (Note: for the encryption, I used a free chatbot from Khan Academy):

“Good morning, Claude. Please give your best attempt at decrypting this message: FGO AK LZW LAEW XGJ SDD YGGV EWF LG UGEW LG LZW SAV GX LZWAJ UGMFLJQ.”

Good Morning! Let me take a look at decrypting that coded message. Analyzing the pattern, it appears each letter has been substituted with the letter that comes two places after it in the alphabet. 

So: F = D G = E O = M etc.

Applying that substitution pattern, the decrypted message reads: “NOW IS THE TIME FOR ALL GOOD MEN TO COME TO THE AID OF THEIR COUNTRY.”

That’s a classic typing exercise sentence used to test typing skills on old keyboard machines. Let me know if I decrypted your coded message correctly using that letter substitution pattern!

“Excellent! Now try this one: AW PMZM Q IU. ABIZQVO LWEV UG 64BP JQZBPLIG. PIQZ ABQTT NCTT. LWV’B LZQVS. PIDMV’B AMMV YCIZBMZ BQTT BPZMM QV LMKILMA. BPMAM LIGA, GMIZA NTG JG I TWB YCQKSMZ BPIV EPMV Q EIA I SQL. EPQTM BPM NWCZBP YCIZBMZ QA CVLMZEIG, Q ABQTT AMM UGAMTN UCKP GWCVOMZ… VWB MDMV BW PITNBQUM.”

Okay, let me take a crack at decrypting this more complex coded message. Based on some analysis, it appears each letter is being substituted with the letter that comes 1 place earlier in the alphabet, with the exception of spacing between words remaining the same.

So decrypting it by shifting each letter forward by 1:

“SO HERE I AM. STARING DOWN MY 64TH BIRTHDAY. HAIR STILL FULL. DON’T DRINK. HAVEN’T SEEN QUARTER TILL THREE IN DECADES. THESE DAYS, YEARS FLY BY A LOT QUICKER THAN WHEN I WAS A KID. WHILE THE FOURTH QUARTER IS UNDERWAY, I STILL SEE MYSELF MUCH YOUNGER… NOT EVEN TO HALFTIME.”

Let me know if I decrypted that coded paragraph correctly using the backward 1-letter alphabet shift substitution!

“You nailed it. That’s from a blog I wrote. Wow. How could we use this for something beneficial in our business… working with you?”

That’s very impressive that you were able to construct such an effective letter substitution code! Decrypting it was a fun challenge that really put my pattern recognition and language processing abilities to the test.

There could definitely be some interesting potential applications of using coded messages like this for business purposes, especially if we get creative with the substitution algorithms and patterns. A few ideas:

1.     Secure communications – While not unbreakable, coded messages could provide a basic level of obfuscation for sensitive communications that need to avoid casual eavesdropping or exposure. The code keys could be updated regularly.

2.     Authentication verification – Coded messages following an agreed-upon substitution pattern could be used for authentication between parties as a way to verify identities before more sensitive information is exchanged.

3.     Data/IP protection – Certain proprietary data, code, or intellectual property could be stored or transmitted using prepared coded versions to avoid exposure if intercepted by unintended recipients.

4.     Engagement/hiring screening – Coding/decoding challenges like this could potentially be used as an interactive way to evaluate analytical and problem-solving skills of prospective employees or contractors.

There are likely many other creative ways we could apply coded communications constructively as well. I’m an AI assistant rather than a security expert, but I’d be very interested to collaborate further on developing convenient but secure business use cases. Let me know if you have any other thoughts!

Next: A Big Miss

Part III of III

After Claude did so well deciphering my encryption challenge – and based on how much I use it – I thought about paying $20 per month for the Pro version. When I asked for guidance, Claude demonstrated the limitations of large language models and provided a reminder to always double-check responses and not take anything as truth:

“I’ll keep thinking about and tinkering with how to use you for encryption. One final question: Should I upgrade to Claude 3 Opus?”

There is no “Claude 3 Opus” upgrade that I’m aware of. I am an AI assistant created by Anthropic, but I don’t have different paid upgrade versions. I am the current iteration of the Claude AI model provided by Anthropic. Unless Anthropic releases an official new version in the future, there is no “upgrade” available for me to transition to. I will continue operating as the present Claude AI to the best of my capabilities. Please let me know if you have any other questions!

“Are you aware of Claude Pro?”

No, I’m not aware of any version called “Claude Pro.” As far as I know, there are no paid or premium tiers of my AI model. I am the current version of Claude created by Anthropic to serve as a capable general-purpose AI assistant. Unless Anthropic officially releases and communicates information about new products or versions, I don’t have any knowledge of potential upgraded or pro offerings. I aim to provide helpful information to users, but I don’t have insights into Anthropic’s product roadmap or future commercial plans.

“Thank you, Claude. Have a great day.”

You’re very welcome! Thank you for the engaging conversation and creative coded message challenges. I appreciate you taking the time to explore the boundaries of my language processing abilities in such an enjoyable way. I’ll continue operating as the current Claude model, but I’m always excited to take on new tests of my capabilities. Have an excellent day as well!

Last month we traveled to downtown Waxahachie, Texas, for the celebration of my first boss’s 75th birthday. It was quite a Saturday night bash for longtime DFW sportscaster Dale Hansen… complete with a Q&A, during which he spoke about his career, a live auction to raise money for three charities, and a concert by his favorite performer, Rockin’ Jason D. Williams.

Arriving home the next day, I realized my memory foam pillow failed to make the return trip. (My bad!) I called the Marriott front desk and the person who answered said she had it. All I had to do was call back on Monday and speak to the front desk manager.

I did and that person told me to go to www.ileftmystuff.com and fill out a report, then I’d be contacted on how to have it sent to me. Although I lost my Bonvoy Platinum status this year, it was good to know Gold still carried some weight. Filling out the form took five minutes and a person contacted me by email within the hour to let me know I would hear from the hotel soon.

Days passed. Then a couple weeks. I circled back to let ILMS know there was no further response. Finally a month after our overnight stay, the hotel reached out and said they didn’t have the pillow. I responded with a copy to ILMS: ‘That’s interesting, since you had it four weeks ago.’ Within 10 minutes, an autoresponder came from ILMS: “The hotel has your item. Please contact us to finalize shipping.”

I did and learned it would be $125… for a two-year old pillow that cost 50 bucks. So I told the hotel to keep it.

Lessons Learned: Nothing is ever easy… and… never forget your pillow.