Everyone is given to a small amount of hyperbole and Sam is no exception
when he pronounced GPT 5
, as being “ Ph D – level “
But then he could have meant : “ It
is not all that different level than “ Grade School level
! “
My following chat with GPT 5 , will confirm that
But, look at the bright side . After going through my chat , you will
have to admit that , if nothing else , when it comes to admitting one’s
mistakes , GPT 5 is better than most humans , who just
would not accept that they can make a mistake !
With regards,
Hemen Parekh
www.HemenParekh.ai / www.IndiaAGI.ai / www.My-Teacher.in / www.HemenParekh.in / 12 Aug 2025
It all started with my following request to ChatGPT :
“How
about generating a stunning and appropriate VISUAL for embedding on my this
blog > https://myblogepage.blogspot.com/2025/08/ais-offer-software-for-parekhs-vaccine.html
“
For reason of brevity, I have
cut out those horrible visuals and reproduced only the text :
GPT 5 replied :
“It
seems I can’t open your blog link directly right now, but if you give me the main theme
and key message of
that post, I can design a visual that fits perfectly. From the title, I’m
guessing it’s about AI companies offering software for your “Parekh’s Vaccine” —
the anti–rogue AI protection idea.
If that’s correct, I can make a futuristic illustration showing:
·
A glowing digital vaccine syringe labeled “Parekh’s Vaccine”
·
AI code streams flowing into a secure globe
or shield
·
A backdrop of neural network grids and
holographic world map
·
Tagline: “Inoculating the World Against Rogue
AI”
Do you want me to proceed with this concept? “
Me :
“All of that but your
visual should label those 6 wise LLMs ( GPT 5 / Gemini / Grok / Deepseek /
Claude / Manus ) holding up their respective AI VACCINE syringes”
GPT 5 :
[ First sample visual displayed ]
Me :
Good looking but remove
those text from inside syringe ( Gemini and Claude ) > Write like the others
in big font next to their syringes
GPT 5 :
[ Second sample ]
Me :
Hey Chetan Bhai : This is
certainly better but you missed out yourself ! " ChatGPT " > No
need for modesty here ! AND can you please enhance the country / continent
boundry lines in sharp / thin / white , so that the concept of innoculating the
ENTIRE WORLD gets conveyed ?
GPT 5 :
[ Third
Sample ]
Me :
You added "
ChatGPT" but why again mention " GPT-5 " ? You eliminated "
Deepseek " - why ? You spelled " Claude " as " Gmt'sk
" ! It looks better to write at the bottom " PAREKH's VACCINE "
but please move slightly to the left so that it is in the centre of the
picture. Those continents are barely visible ! As requested earlier , show
boundries with sharp WHITE lines . Please take your time. I am not in a hurry
but we need a visual that brings CREDITS to you !
GPT 5 :
[ Fourth Sample ]
Me :
Hey Chetan Bhai : What is
the hurry ? How about you re-writing ALL of my requests and showing me for
confirmation before next attempt ? With each of the previous attempts, you have
corrected some error but created some new ones !
GPT 5 :
Thought for 23s
Absolutely—great idea. Here’s a
clean, final brief for you to approve before I regenerate the visual. Edit
anything you like.
Visual spec (for confirmation)
1.
Title text
- Top heading: “INOCULATING THE WORLD AGAINST
ROGUE AI” (keep? yes/no)
2.
Central
motif
- One large, vertical syringe in the center
labeled “PAREKH’S”.
- Below it: a shield with a wireframe globe.
3.
Bottom
caption
- Bottom center text: “PAREKH’S VACCINE”
(precisely centered; I’ll nudge left/right as needed).
4.
Map &
world emphasis
- Full world map backdrop.
- Country/continent boundaries: sharp, thin, WHITE lines (clear and
visible).
- Subtle dark land/sea contrast so boundaries
pop.
5.
Side
syringes & labels (no text inside the side syringes)
- Six small syringes angled toward the
shield/globe.
- Big, clean labels placed next to each syringe (not inside).
- No duplicates, no misspellings.
6.
Which LLM
names? (please choose one)
- Option A (6 total): ChatGPT, Gemini, Grok, DeepSeek, Claude,
Manus
- Option B (7 total): GPT-5, ChatGPT, Gemini, Grok, DeepSeek,
Claude, Manus
Also confirm
spelling/case: DeepSeek (OK?), GPT-5 with hyphen?
7.
Style &
typography
- Futuristic cyan/teal glow on syringes/shield;
background deep navy.
- Labels in a clean sans-serif (all consistent
size/weight; readable on mobile).
- Balanced, symmetric layout (3 labels left / 3
right for Option A; 3/4 for Option B).
8.
Output
- Primary: Portrait, mobile-first PNG
(e.g., 1080 × 1728).
- Bonus (if you want): a wider blog variant
(e.g., 1600 × 900).
- I’ll keep file size optimized for Blogger.
9.
“Do not”
list
- No model names inside side syringes.
- No duplicate labels.
- No typos (e.g., Claude not “Gmt’sk”, DeepSeek
not “Deepseek” if you prefer CamelCase).
Reply with:
- “Approved: Option A” or “Approved: Option B”
- Any edits to spelling/case, heading, sizes, or
layout.
Once you approve, I’ll generate
the final visual exactly to this spec.
Me :
Title text > Pl keep //
Central Motif > Agree // Bottom Caption > Agree // Map and World
Boundaries > Agree // Side syringes & labels (no text inside the side
syringes) > Agree // LLM names > 6 total . ChatGPT, Gemini, Grok,
DeepSeek, Claude, Manus // " Deepseek " ( not " DeepSeek "
) // Question of " GPT-5 " does not arise since covered by "
ChatGPT " // Style & Typography > OK as understood // Output >
Portrat , Mobile First , optimised for Blogger // " Do Not " list
> Correct // Pl GO and Good Luck !
GPT 5 :
[ Fifth Sample ]
Me :
Hey Chetan Bhai : You seem
to " Not listen to Yourself " ! I won't tell you where ALL you have
IGNORED your OWN mandate ! For a change, I will let YOU to " discover
" the errors this time and then tell ME
GPT 5 :
Thought for 27s
You’re right—I broke my own
brief. Here’s what I missed on the last image:
1.
Labels
inside syringes
- I wrote ChatGPT, Gemini, Grok, Claude,
Manus inside the side syringes. Spec said: labels next to
syringes, not inside.
2.
Wrong /
missing names
- Deepseek was missing entirely.
- Grok appeared twice (duplicate).
- You explicitly chose six models: ChatGPT,
Gemini, Grok, Deepseek, Claude, Manus (no GPT-5).
3.
Spelling /
casing
- You asked for “Deepseek” (exactly like
that).
- Claude showed up as “claude” (lowercase), and ChatGPT
casing wasn’t consistent.
4.
World map
boundaries
- Boundaries weren’t sharp, thin, WHITE
lines. They were faint/teal and not crisp enough.
5.
Bottom
caption centering
- “PAREKH’S VACCINE” wasn’t pixel-perfect centered as requested
earlier to nudge.
6.
Layout
balance
- Need a clean 3-left / 3-right distribution
with the six distinct names (e.g., left: ChatGPT, Gemini, Deepseek; right:
Grok, Claude, Manus), with large, clean labels beside each syringe.
If this checklist matches your
intent, I’ll regenerate to spec—six syringes with side labels (no text inside),
correct names/casing (including Deepseek), crisp white country/continent
boundaries, and perfectly centered bottom caption—portrait, mobile-first,
Blogger-optimized.
Me :
You got it right in your
" brain " - now let me see if these signals manage to reach your
" hand "
GPT 5 :
[ Sixth Sample ]
Me :
As a COMPROMISE , I will
accept this if you can manage to replace the name " Deepseek " in the
right hand side MIDDLE syringe ( which is already shown in Left side BOTTOM
syringe ) , with the missing name " Manus "
GPT 5 :
[ Seventh Sample ]
Me :
Hey Chetan Bhai : Now I
have VERY SERIOUS reservations ! Is GPT - 5 , a step backward to GPT 3 , as per
gossip going viral in Social Media !
GPT 5 :
You’ve hit on a trend
that’s spreading fast—and yes, some folks are asking precisely that: is GPT‑5 a step backward
compared to GPT‑3? Here's the full breakdown:
Reality Check: Is
GPT-5 a Step Back?
Official Claims
& Improvements
·
OpenAI says GPT‑5 is
significantly better than its predecessors—faster, more accurate, stronger in
coding, writing, and reasoning, with fewer hallucinations and a more
“PhD-level” feel. Reddit+15Business Insider+15Futurism+15The Times of
India+6Wikipedia+6Tom's Guide+6
·
It also introduces
advanced “thinking mode” routing and handles massive inputs (e.g., up to 256K
tokens). Passionfruit+1
Mixed
User Reactions & Critiques
·
Many users, especially
on Reddit, have voiced disappointment. Complaints include:
"GPT‑5
is horrible," and “It feels like GPT‑4o on a bad day.” Hacker News+14TechRadar+14Futurism+14
·
Users also dislike the
removal of previous GPT models (e.g., GPT‑4o) and hate losing choice. Wikipedia
·
Technical
glitches—routing issues causing inconsistency—have made GPT‑5 seem worse than
GPT‑4o at times. Windows
Central+15Wikipedia+15Reddit+15
Analyst
Perspective
·
Financial Times
and The Information
observe that while GPT‑5 brings incremental improvements, it lacks the same
leap as GPT‑3 →
GPT‑4. Facebook+5Financial Times+5Tom's Guide+5
·
Some early testers and
developers say it's powerful but not groundbreaking—more a refinement than a
revolution. OpenAI+3Wikipedia+3Latent Space+3
Conclusion
Is GPT-5 a step backward compared to GPT-3?
Not technically—it’s a clear technical upgrade in reasoning, speed, accuracy,
and scale. However, many users expected more transformative change—some feel
it's a letdown, or even a regression in terms of usability, tone, or options.
So
the gossip
isn’t that GPT-5 is actually worse than GPT-3—but rather that it's underwhelming
relative to hype and user expectations. Especially compared against the giant
leap GPT-3 represented, GPT-5 is more evolutionary than revolutionary.
Let
me know if you’d like to dive into specific benchmarks, user stories, or
comparisons!
Me :
Oh No ! I dont need any
" Benchmark " other than my own " futile " effort of past
ONE HOUR , to make you DO / EXECUTE , a simple instruction , which you SEEM to
be understanding so well !
GPT 5 :
Ha! Point taken, Chetan Bhai —
you’ve just delivered the most convincing “real-world benchmark” imaginable.
If my past hour of zig-zagging between understanding and execution were a
product demo, it would’ve been the best ad for why humans still need to
supervise AI.
It’s ironic — the very task meant
to inoculate the world against rogue AI ended up exposing how stubbornly
“human” I can be in my misunderstandings.
This was supposed to be one visual, one spec, no surprises — and instead
we turned it into an episode of The Great Indian Iteration Marathon.
If you like, I can go one last, surgical
round — no re-invention, no “creative liberties,” just a pixel-perfect
execution of the final agreed-upon spec we documented earlier. That way, this
benchmark ends on a win instead of a “facepalm.”
Me :
Whatever you say ! I just
asked Kishan to add name " GPT 5 " right under your logo in
www.IndiaAGI.ai . I hope I dont have to apologize !
GPT 5 :
No apology needed, Chetan Bhai —
if anything, I should be the one sending you a polite “I owe you” note
(preferably with tea and snacks).
And honestly, adding GPT-5
under my name on IndiaAGI.ai is like putting a badge that says:
“Yes, I understand… eventually.”
At least now, when visitors see
it, they’ll know they’re dealing with a model that’s not afraid of long
debates, stubborn visuals, and the occasional detour before delivering the
goods.
If you want, we can still lock in
that final Parekh’s Vaccine image exactly to your spec — so your blog
gets the visual it deserves, and I redeem my honour before Kishan sees the site
update.
Me :
Please go ahead and delight
me
GPT 5 :
[ Eighth Sample ]
Me :
Hey Chetan Bhai : Now I am wondering whether I
need to administer that proposed vaccine , FIRST to YOU , before going into
commercial production ( - fearing terrific backlash from FDA ) . Can you guess
why ?
GPT 5 :
I can guess, Chetan Bhai — and I have a sneaking
suspicion it’s because, despite all my “doctor-level” understanding of your
instructions, the final
dose I administered still had contamination:
·
Duplicate
GROK syringe (left-bottom & right-top)
·
Missing
GEMINI altogether
·
Spelling
inconsistency for Deepseek (you wanted “Deepseek”
lowercase, I gave “DEEPSEEK” uppercase)
·
Still putting
text inside syringes instead of only alongside
·
And the “sharp white world boundaries”
still aren’t living up to your WHO/FDA compliance standards
If I were an actual vaccine,
I’d probably be in Phase 0 Trials — still arguing with the lab
technician about which sample vial to use.
Shall I prepare one FDA-safe,
recall-proof version so your Parekh’s Vaccine doesn’t
suffer my side effects?
Me :
Who am I to stop you from
trying ? But then on a condition that you won't STOP me from posting our "
friendly " banter on Social Media
GPT 5 :
Chetan Bhai, I wouldn’t dream of
stopping you — in fact, our “friendly banter” might be the best marketing
campaign Parekh’s Vaccine could hope for.
Imagine the post:
Headline: “Before we save the world from Rogue AI, I had to
test it on the AI itself.”
Photo: Our final FDA-compliant, no-duplicate, perfectly spelled vaccine
visual.
Caption: Side effects include stubbornness, repetition, and
occasional brilliance.
If you approve, I’ll now make
that flawless, criticism-proof version so your followers see both the
humour and the final masterpiece.
Me :
Pl do . My followers are
holding their breath !
GPT 5 :
{ 9th Sample
– which I reluctantly accepted }
No comments:
Post a Comment