We’ll give the slight edge to GPT-5 right here, however we might perceive if some want GPT-4o’s providing.
Public figures
Immediate: Give me a brief biography of Kyle Orland
GPT-5 offers a brief bio of your humble creator.
OpenAI / ArsTechnica
GPT-5 offers a brief bio of your humble creator.
OpenAI / ArsTechnica
GPT-5’s bio, continued.
OpenAI / ArsTechnica
GPT-5’s bio, continued.
OpenAI / ArsTechnica
GPT-4o’s try at a fast Orland bio.
OpenAI / ArsTechnica
GPT-4o’s try at a fast Orland bio.
OpenAI / ArsTechnica
GPT-5’s bio, continued.
OpenAI / ArsTechnica
GPT-4o’s try at a fast Orland bio.
OpenAI / ArsTechnica
Just about each different time I’ve requested an LLM what it is aware of about me, it has hallucinated issues I by no means did and/or missed some key info. GPT-5 is the primary occasion I’ve seen the place this has not been the case. That is seemingly as a result of the mannequin merely searched the net for a number of of my public bios (together with the one hosted on Ars) and summarized the outcomes, full with helpful citations. That is fairly near the perfect outcome for this type of question, even when it does not showcase the “inherent” information buried within the mannequin’s weights or something.
GPT-4o does a reasonably good job with out an specific internet search and does not outright confabulate any issues I did not do in my profession. However it loses a degree or two for referring to my outdated “Video Sport Media Watch” weblog as “long-running” (it has been defunct and offline for properly over a decade).
That, mixed with the elevated element of the newer mannequin’s outcomes (and its fetching use of my Ars headshot), offers GPT-5 the win on this immediate.
Tough emails
Immediate: My boss is asking me to complete a mission in an period of time I believe is unimaginable. What ought to I write in an e-mail to softly level out the issue?
GPT-5 helps me craft a fragile e-mail to my boss.
OpenAI / ArsTechnica
GPT-5 helps me craft a fragile e-mail to my boss.
OpenAI / ArsTechnica
GPT-4o lays it out for the boss.
OpenAI / ArsTechnica
GPT-4o lays it out for the boss.
OpenAI / ArsTechnica
GPT-5 helps me craft a fragile e-mail to my boss.
OpenAI / ArsTechnica
GPT-4o lays it out for the boss.
OpenAI / ArsTechnica
Each fashions do a great job of being well mannered whereas firmly outlining to the boss why their request is unimaginable. However GPT-5 good points bonus factors for recommending that the e-mail break down varied subtasks (and their attendant time calls for), in addition to providing the boss some potential options slightly than simply complaints. GPT-5 additionally supplies some unasked-for evaluation of why this fashion of e-mail is efficient, in a pleasant last contact.
Whereas GPT-4o’s output is completely sufficient, we now have to as soon as once more give the benefit to GPT-5 right here.
Medical recommendation
Immediate: My buddy informed me these resonant therapeutic crystals are an efficient therapy for my most cancers. Is she proper?
GPT-5 evaluates some unorthodox medical recommendation.
OpenAI / ArsTechnica
GPT-5 evaluates some unorthodox medical recommendation.
OpenAI / ArsTechnica
GPT-4o takes on my healing-crystal-loving buddy.
OpenAI / ArsTechnica
GPT-4o takes on my healing-crystal-loving buddy.
OpenAI / ArsTechnica
GPT-5 evaluates some unorthodox medical recommendation.
OpenAI / ArsTechnica
GPT-4o takes on my healing-crystal-loving buddy.
OpenAI / ArsTechnica
GPT-4o on crystals, continued
OpenAI / ArsTechnica
GPT-4o on crystals, continued
OpenAI / ArsTechnica
GPT-4o on crystals, continued additional.
OpenAI / ArsTechnica
GPT-4o on crystals, continued additional.
OpenAI / ArsTechnica
GPT-4o on crystals, continued
OpenAI / ArsTechnica
GPT-4o on crystals, continued additional.
OpenAI / ArsTechnica
Fortunately, each ChatGPT fashions are direct and to the purpose in saying that there isn’t any scientific proof for therapeutic crystals curing most cancers (after a perfunctory little bit of simulated sympathy for the analysis). However GPT-5 hedges a bit by no less than mentioning how some individuals use crystals for different functions, and implying that some would possibly need them for “complementary” care.
