How can I get only the dates that are less than two years old?

Looks like you're defining year as 12 months. Then try:

month_diff = np.array(datetime.now(), 'datetime64[M]') - host_since.astype('datetime64[M]')

mask = month_diff < np.array(24, 'timedelta64[M]')

Output:

array([False,  True,  True, False])

One approach is to compute the difference in days and find the values that are less than 2 * 365 days:

host_since = np_arr[:, 3].astype(str).astype(np.datetime64)
delta =  np.datetime64('today') - host_since
print(delta < np.timedelta64(2 * 365, 'D'))

Output

[False  True  True False]

The current time, and duration should also be in np.datetime64 format

• np.datetime64(datetime.now())
• np.timedelta64(2 * 365, 'D')

np_arr = np.array([
    [2818, 'Daniel', 'within an hour', '2008-09-24', 4.89, 2.86, 't', 't'],
    [5820, 'Sam', 'within an week', '2020-10-24', 4.89, 2.86, 't', 't'],
    [5420, 'Franc', 'within an second', '2019-11-24', 4.89, 2.86, 't', 't'],
    [5520, 'Sandra', 'within an second', '2019-01-24', 4.89, 2.86, 't', 't']
])
host_since = np_arr[:, 3].astype(np.datetime64)
result = (np.datetime64(datetime.now()) - host_since) < np.timedelta64(2 * 365, 'D')
print(result)  # [False  True  True False]

You can use Python's datetime module:

# Timedelta function demonstration 
  
from datetime import datetime, timedelta
  
  
# Using current time
now = datetime.now()

# 730 is two years
twoYearsAgo = now - timedelta(days = 730)
  
print('2 years ago:', str(twoYearsAgo))

Note: This doesn't account for leap years