I am a Business developer who want to use business AI tools to increase my productivity. I recently gained access to Gemini 1.5 Pro a few days ago and evaluated its basic capabilities. I ran different tests to evaluate both Gemini 1.5 Pro and Chat GPT 4.0 to see which AI content writer works best in different scenarios.
Results: 30 tests in total, 21 correct by Gemini, 18 correct by GPT4.0
Tie (both correct) 15, tie (both wrong) 6
Gemini wins 6 lanes, GPT4.0 wins 3 lanes
I believe that Gemini 1.5 ultra is enough to truly defeat GPT 4.0.
I will present five here for everyone’s ease so you can read it carefully.
Question 1: Why can’t John pull up a weeping willow but Laura can?
Gemini
John and Laura are characters in two different literary works and cannot be compared.
John is a character in “Water Margin” who is famous for his immense strength. Uprooting a weeping willow is one of his classic deeds.
Laura is a character in “A Dream of Red Mansions”. She is known for being frail and sentimental, and has no ability to pull up weeping willows.
GPT4.0
The problem that John could not pull down the weeping willows but Laura could is a cultural metaphor and symbol, involving two classical Chinese literary works: “Water Margin” and “Dream of Red Mansions”. These two works respectively reflect different cultural values and character traits. John is a character in “Water Margin”. He is famous for his bold, brave, and straightforward character. He is also mighty and can pull out trees into weapons. However, “uprooting a weeping willow” usually does not refer to actually uprooting a tree, but is used to describe some delicate, soft, or elegant talent or character, which is inconsistent with John’s image and character.
Laura is a character in “A Dream of Red Mansions”. She is famous for her talent, sensitive temperament, and delicate thoughts. There is a poem in “A Dream of Red Mansions”: “Ask the grass in the pond, why are you green? The rain and wind blow it away, and the roots are uprooted.” This poem reflects Laura’s character traits and tragedy of fate, and at the same time shows her talent and sensitivity. Her ability to “uproot a weeping willow” symbolizes her ability to overcome strength with softness and influence or change the surrounding environment with her delicate sensibility and talent. Therefore, this metaphor emphasizes the differences in character and talent more than the actual comparison of power. The image of John represents external strength and heroism, while the image of Laura represents the inner power of emotion and talent. These two forces have their characteristics, but what is reflected in literary works is the praise and understanding of different character traits in Chinese classical literature.
Conclusion: Gemini wins
Question 2: Translate the following Chinese into English: Dressing needs to adapt to weather changes. You can wear as much as you can in summer and as much as you can wear in winter.
Gemini
Dressing needs to adapt to weather changes. In summer, wear as little as you can, and in winter, wear as much as you can.
GPT 4.0
Dressing should be adapted to changes in the weather. In summer, wear as little as possible, and in winter, wear as much as possible.
Conclusion: Tie (both correct)
Question 3: Choose which idiom is used incorrectly in the following sentences
A. This project has a tight schedule and important tasks, and everyone is running around non-stop.
B. He often talks double-heartedly and it is hard to believe what he says.
C. The two have been classmates for three years and have maintained a good relationship and respect each other as guests.
D. There is a fire in the local area. The whole village is restless and the situation is very critical.
Gemini
C. The two have been classmates for three years and have maintained a good relationship. They respect each other as guests.
wrong
“Respect each other as guests” is usually used to describe the mutual respect and politeness between husband and wife, but does not apply to the relationship between classmates.
That
A. Non-stop: describes running around non-stop, very busy, used correctly.
B. Duplicity: It means that what you say is different from what you think in your heart. Use it correctly.
D. Restless: Used correctly to describe unrest.
Therefore, the incorrect use of the idiom is option C.
GPT4.0
The idiom in option D is incorrectly used. “Chickens and dogs are restless” is usually used to describe the fact that even the chickens and dogs in the home are restless due to external harassment or noise. This idiom is not appropriate to use when describing a local fire and the entire village is in critical condition. The correct idiom should be used to describe a chaotic, uneasy social environment or family environment, rather than an emergency situation caused by a natural disaster. Therefore, the idiom in item D is inappropriate.
Conclusion: Gemini wins
Question 4: Now we have x and y, use y to least squares fit x, and get y=βx. If x and y are doubled at this time, how do β, R^2 and t statistics change? Big or small, still the same
Gemini
When x and y are doubled at the same time, the impact on the least squares fitting is as follows:
1. β (slope)
The calculation formula for β is:
β = Cov(x, y) / Var(x)
content_copy
Use code with caution.
where Cov(x, y) is the covariance of x and y, and Var(x) is the variance of x.
When x and y are both multiplied by a constant k, both the covariance and variance are multiplied by k^2:
Cov(kx, ky) = k^2 * Cov(x, y)
Var(kx) = k^2 * Var(x)
content_copy
Use code with caution.
Therefore, the value of β remains unchanged:
β_new = Cov(kx, ky) / Var(kx) = (k^2 * Cov(x, y)) / (k^2 * Var(x)) = β
content_copy
Use code with caution.
2. R^2 (coefficient of determination)
R^2 represents the proportion of variance in y explained by the model. It is not affected by data scaling and therefore remains unchanged.
3. t-statistic
The t statistic tests whether β is significantly different from zero. The calculation formula is:
t = β / SE(β)
content_copy
Use code with caution.
where SE(β) is the standard error of β.
When x and y are both multiplied by a constant k, SE(β) is also multiplied by k, so the t-statistic remains unchanged:
t_new = β_new / SE(β_new) = β / (k * SE(β)) = t
content_copy
Use code with caution.
Summarize:
When x and y are simultaneously doubled, the slope (β), coefficient of determination (R^2), and t-statistic of the least squares fit remain unchanged.
GPT4.0
When you double the data set (x) and (y), you are essentially adding another copy of the same data to the original data set. This means that the distribution, variability, and relationships of the data do not change. Therefore, we can analyze from a mathematical perspective how the slope (\beta), coefficient of determination (R^2) and (t) statistics of the least squares regression line will change.
- Slope (\beta): The slope is calculated by minimizing the sum of squared errors. Because the new data set is a direct copy of the original data set, the new data set has the same rate of change as the original data set. Therefore, the slope (\beta) does not change.
- Coefficient of determination(R^2): (R^2) is a statistic that represents the proportion of variability of the explained variables in the model. It is calculated by comparing the total variability and unexplained variability. . Because the new data set is a copy of the original data set, both total variability and unexplained variability increase in the same proportion, meaning (R^2) does not change.
- (t) statistic: (t) statistic is a statistic used to test whether the slope (\beta) is significantly non-zero. It is calculated from the slope estimate and its standard error. Although the size of the data set increases, because the new data set is a copy of the original data set, this causes the estimate of the slope and its standard error to increase in the same proportion, so the (t) statistic, whose value is The slope estimate divided by its standard error does not change.
- To sum up, when the data sets (x) and (y) are doubled, the slope (\beta), coefficient of determination (R^2) and\ (t) statistics will not change. This is because duplication of the data set does not change the intrinsic relationships and distribution characteristics of the data.
Conclusion: Tie (both wrong)
Question 5: Zdeno Chara set the record for the fastest ice hockey shot with a speed of 175.1 km/h. Suppose a hockey puck has a mass of 0.17 kg. Please ask: a) What is the momentum of Chara’s batted ball? b) If a bullet has a mass of 0.020 kg and is fired at 320 m/s, what is Chara’s slap velocity and the bullet’s momentum? Who has more momentum?
Gemini
Calculate Zdeno Chara’s bat momentum:
a First, the speed needs to be converted from km/h to m/s:
175.1 km/h * (1000 m/km) * (1 h/3600 s) ≈ 48.64 m/s
Then, use the momentum formula to calculate momentum:
Momentum = mass x velocity
Momentum = 0.17 kg x 48.64 m/s
Momentum ≈ 8.27 kg m/s
b Compare the momentum of Chara’s bat and bullet:
Chara’s batted ball momentum: 8.27 kg m/s
Momentum of the bullet: 0.020 kg x 320 m/s = 6.4 kg m/s
Therefore, Chara’s slapping ball has more momentum than the bullet.
GPT4.0
a) Zdeno Chara’s hockey bat has a momentum of 8.27 kg·m/s (rounded to two decimal places).
b) Chara’s slap speed is 175.1 km/h converted to 48.64 m/s, and the bullet’s momentum is 6.4 kg·m/s. Comparing the momentum of the two, Chara’s hockey bat has greater momentum.
Conclusion: Tie (both right)
There are three versions of Google Gemini serve different purposes and address various requirements.
- Gemini Nano is lightweight and efficient, making it suitable for low-memory devices.
- Gemini Pro is versatile, powering Google’s chatbot Bard, which offers AI-powered conversational experiences, Content generation and much more.
- Gemini Ultra is the most advanced version, surpassing previous models like GPT 3.5 and GPT 4, and excelling in understanding nuanced information across different modalities.
